I'm not a fan of tank-to-tank testing. Still too many variables (weather, being the biggest, but not only one).
See:
http://ecomodder.com/forum/showthrea...ery-11445.html
I'd be more inclined to do A-B-A coastdown testing, where you're only looking for a relative difference in coastdown time/distance in relatively controlled circumstances, rather than trying to measure MPG change in the real world.