Just thought folks would like to see these E-Trac numbers. I tested 33 war nickels and the results varied quite a bit.
1 read 12-14
5 read 12-15
12 read 12-16
10 read 12-17
4 read 12-18
1 read 12-19

I know air testing is biased, but knowing the range these vary by may be helpful. Unfortunately these numbers run directly into pulltab territory.

As long as I was setup to test, I grabbed some other coins too:

Dimes
13-45 -- 7 coins (Barbers and teens Mercs)
13-46 -- 30 coins
13-47 -- 87 coins (only one Barber)
14-45 -- 1 coin (Barber)
14-46 -- 33 coins
14-47 -- 4 coins

Quarters
06-48 -- 1 coin (Washington)
07-48 -- 8 coins
08-48 -- 16 coins
09-48 -- 4 coins (SLQ and Barber)
10-48 -- 1 coin (SLQ)

Halves
03-48 -- 5 coins
05-48 -- 2 coins
07-48 -- 2 coin (Barber)
05-49 -- 1 coin (Barber)

I am curious as to why dug indian heads and early wheats come out of the ground with such nice green patina.  It's too much for me to believe this is just chance that coins after the mid-1920's do not exhibit this patination. Two cent pieces minted from 1864 to 1872 also have these same characteristics.
 
I believe the answer lies in the "5% tin and zinc" content of bronze cents.  I cannot find a reference to the specific percentages of zinc and tin used.  Sometime in the early 1920s, the percentage of zinc must have been increased and tin decreased.  I can find no reference to this change, but I would be very interested in finding one.

Tin is a key element in bronze.  The properties of tin state that it is not easily oxidized, prevents corrosion, resists corrosion from water, but can be attacked by acids, alkalis and acid salts.  These are exactly the properties that would make the early cents better preserved than the later ones.

Anyone who has dug an eaten up modern penny knows how zinc weathers in the ground.  So, this also indicates that an increase in zinc would make the surfaces more easily corroded.

Follow up #1
The composition is the "same" only to the extent that it is "5% tin and zinc" this is a vague reference.  I can find no references that specify the specific percentages of tin and zinc.

My theory is the older "green" cents are maybe 3% tin and 2% zinc and the non-green are maybe 1% tin and 4% zinc.

It has little to do with the time in the ground, as these earlier cents even read differently on the meter of many metal detectors, which says to me that it is clear there is some difference in metallic content.  The conductive value of older cents on the Minelab E-Trac is 33-36, the conductive value of later cents is 42-43.  And this can be shown in the ground and in air tests with dug and non-dug coins, so the time in the ground is not significant.

Follow up #2
Well, the minerals or ground effect only prevent the green patination from forming or eat it away.  If the soil is "acid, alkali or contains acid salts" then the tin will be attacked and the coin will have a rough corroded surface.  

Where the soil is fairly neutral, the tin provides some protection against corrosion.  There is some oxidation/corrosion occurring in the ground as that is what makes the cent turn green.  

My hypothesis is a greater amount of tin in the mix helps keep the surfaces smooth and even, preventing (in more neutral soils) the rough or heavy corrosion as seen in many copper coins recovered from fertilized farm fields.

I find that in my soil, which is fairly stable and neutral that older small cents tend to have this smooth patination.  Coins that have been fertilized or are in clay type soil do not.  Even modern coins in clay type soil are heavy corroded.

Overall I guess my question is not, "Why do they turn green?" but rather, why are the early cents so much better preserved than later cents when supposedly the metals are the same.  I am saying is the metals must not really be the same, and the ratios of tin to zinc within that 5% have been altered.

Another random thought is wondering if the addition of tin to the copper coating on modern zinc pennies would improve their corrosion resistance. Hmm...

Follow up #3
Wikipedia shed some light on some things:

Brass is any alloy of copper and zinc; the proportions of zinc and copper can be varied to create a range of brasses with varying properties. In comparison, bronze is principally an alloy of copper and tin.

Aluminium makes brass stronger and more corrosion resistant. Aluminium also causes a highly beneficial hard layer of aluminium oxide (Al2O3) to be formed on the surface that is thin, transparent and self healing. Tin has a similar effect and finds its use especially in sea water applications (naval brasses). Combinations of iron, aluminium, silicon and manganese make brass wear and tear resistant. A well known alloy used in the automotive industry is 'LDM C673', where the combination of manganese and silicon leads to a strong and resistant brass.

Selective leaching, also called dealloying, demetalification, parting and selective corrosion, is a corrosion type in some solid solution alloys, when in suitable conditions a component of the alloys is preferentially leached from the material. The less noble metal is removed from the alloy by microscopic-scale galvanic corrosion mechanism. The most susceptible alloys are the ones containing metals with high distance between each other in the galvanic series, eg. copper and zinc in brass.

Follow up #4
I discovered this text from Jevon's 1875 book, Money and the Mechanism of Exchange:"

Bronze Coin.
XI.11  It was known, even in prehistoric times, that a small quantity of tin communicated hardness to copper, and the ancient nations were familiar with the use of bronze thus manufactured. The French Revolutionary Government melted up the bells of the churches seized by them, and the sous de cloche,as they were called, made from the bell metal, were superior to coins of pure copper. Yet curiously enough no modern government thought of employing a well-chosen bronze for small money, until the government of the late Emperor of the French undertook the recoinage of the old sous in 1852. This recoinage was carried out with great success.

XI.12  Between the years 1853 and 1867 coins to the nominal value of about two millions sterling, consisting of 800 millions of pieces, and weighing eleven millions of kilograms (10,826 tons) were struck, in addition to a subsequent issue of about 200 millions of pieces. The experiment was in almost every way successful. The ten and five-centime pieces now circulating in France are models of good minting, with a low but sharp and clear impression. They were readily accepted by the people, although only weighing as much as the sous rejected in the time of the Revolution, namely, one gram per centime, and they are wearing well.

XI.13  The bronze used consists of 95 parts of copper, four of tin, and one of zinc. It is much harder than copper, yet so tough and impressible that it takes a fine impression from the dies, and retains it for a long time. It cannot be struck except by a press of some power, and thus counterfeiting is rendered almost impossible. It can hardly be said to corrode by exposure to air or damp, and merely acquires a natural patina, or thin dark film of copper oxide, which throws the worn parts of the design into relief, and increases the beauty of the coin.

XI.14  Bronze has since been coined by the governments of England, the United States, Italy, and Sweden, and it seems probable that it will entirely take the place of copper. The German government is now using bronze for the one-pfennig pieces."

My conclusions drawn from this: Bronze for coins was a new idea in 1852.  The specific formula is listed as containing 4% tin and 1% zinc.  This is most likely the formula used in the first Indian head cents in 1864, as this new bronze alloy would have proven itself with use in Europe.  This supports my contention that the earlier cents have a higher (4%) tin content, and that is why they corrode less and maintain smooth green surfaces.  We know that the tin was removed from cents completely in 1962, but I contend the amount of tin was reduced before 1962.

Domestic and foreign coins manufactured by mints of the United States (1965) states: "Alloy of 95% copper and 5% zinc coined 1944-46 by order of Secretary of the Treasury of Dec. 16, 1943, pursuant to act of Dec. 18, 1942."

According to the same document above, the US mint was producing bronze coins for other countries. In 1917-1919 the US mint produced coins for containing 95% copper, 3% tin and 2% zinc for Peru. In the early twenties, coins were minted containing 95% copper, 1% tin and 4% zinc as early as 1921 (Indo-China) and 1922 (Costa Rica).  In 1942, coins made for Surinam were 95% copper, 1% tin and 4% zinc.   It would seem odd that they would create a unique alloy for these small runs of foreign coins rather than use the "stock" US bronze alloy.

This points to the late teens as a target timeframe that tin was first reduced from the bronze cent, giving a speculative timetable that looks like this:
1864-1916   95% copper, 4% tin, 1% zinc
1917-1920   95% copper, 3% tin, 2% zinc
1921-1942   95% copper, 1% tin, 4% zinc*
1944-1946   95% copper, 5% zinc
1947-1962   95% copper, 1% tin, 4% zinc
1963-1982   95% copper, 5% zinc

*In testing wheat cents with a metal detector, the majority of coins having a arbitrary numerical CO value of 36-41 or were minted before 1942.  Almost all coins minted 1944 and after had CO values of 42-44.  From this interpretation the early 1920s coins may have been somewhat experimental or perhaps normal variations in batch consistency, and the reduction to 1% tin did not occur until 1947.

Follow up #5 (October 1, 2013)

I discovered a reference to the composition of the two cent pieces which exhibit similar characteristics to the early small cents. 'In a Dec. 8, 1863 letter from Mint Director James Pollock to Secretary of the Treasury Salmon P. Chase, the Mint director recommended a 2-cent coin be
introduced in what was called “French bronze.”'

According to wikipedia, French bronze is a form of bronze typically consisting of 91% copper, 2% tin, 6% zinc, and 1% lead.

First portion of The Coinage Act of April 22, 1864