I'd be interested to see more acceleration testing in an a-b-a type experiment. There are just too many variables to consider when comparing tank to tank. In this graph of fill-ups in one of our cars it is super common to see 10+ mpg fluctuations from tank to tank.
