It might mean that the coin is weighted. Sure, the single trial of 20 flips isn't a perfect confirmation, but it's enough to make you suspicious. (There's only an approximate one in a million chance of that happening; 1 in 1,048,576 to be exact.)
If you have a table in the middle of your room, according to quantum mechanics, there is a finite, nonzero probability that the next time you observe this table, all its valence electrons would spontaneously excite, and you would observe a massive combustive photoelectric process for an object that isn't even a metal. However, the probability is so tiny that you can't expect this to ever happen in the age of the universe. If it does, you can probably assume that something is amiss, and that it isn't just some freak occurrence of quantum probability.
The same is true here. There are proper, well established statistical tests that can be used to verify whether or not something is adhering to the probability distribution that you expect, even if it has inherently random outcomes.
I'm not saying that such a test has ever been successfully carried out for D3 drop rates (and, in fact, I don't actually believe that there's anything wrong with them). I'm just illustrating the fact that you can still develop positive suspicions, and even debunk expected distributions from systems by observing a significant enough anomaly, even though there was a finite probability for that anomaly.
Edited by Bottle#1907 on 11/17/2012 7:20 AM PST