I don't have a satisfying technical answer for you but this isn't the first car where I've recorded evidence of a FE hit way above that 3%. I tried the same thing in my Prius over the course of a few months while i was living in SC and the results were similar. The thing to keep in mind is that the 3% figure is based on the raw energy content of fuel. (I.e., E10 theoretically contains 3% less chemical energy than E0.) But what matters is how much mechanical energy an engine can extract from fuel. If you stack a 3% hit on chemical energy on top of another few percent due to some finicky thing like OEM ignition maps being optimized for E0 while E10 would need a couple degrees more advance to be equally efficient, 7% comes pretty quick. There's also the fact that ethanol loves water and will readily absorb it from any available source. What's to stop condensation from forming in a cool underground tank as warm surface air fills the space left as that tank empties? Even a small quantity of moisture in fuel is bound to impact combustion even if we're just talking about a tiny mass fraction. I don't think any cheating or bias is required for these results to be plausible.