This is one case where logical knowledge beats empirical knowledge.

Logical on what criteria? Reasons may be as small and minor as the minute differences in the stamping/rounded corners/age of coin (and we are talking coins not geometric theoretical shapes)/weight of head versus reverse/grease/detrirus/circularity/reeding/rim - other minutia are available.

all practical differences. Empirical is as good as the stated object, and the apparatus that was used to flip (not human interference)

And any conman/prestidigitator (or prestigititrix) could consistently engineer a result and only need a bias in their favour. I am told certain croupiers can hit a given number on a roulette wheel on a good day. They only have to have a hit rate that gives the bank an advantage more than a zero or double zero, to earn their stipend. But that is human, not machine as in the research.