I didn't think of this, but since byte-swapping is so fast (a few seconds at most), I don't think it affects the results of the new longer test significantly -- other system factors seems to change the results by 5% from trial to trial, which I think is a larger effect.