To see wind condition on test i use this web page:
History : Weather Underground
And I do test on the same road on both directions.
I think tuft testing without wheel covers, and rear skirts will be very interesting.
I also think the tuft are too long and vibrations is greater than real.
You could get better accuracy going faster (because higher drag resistance contribution to fuel consumption) at cruise speed, Reset SG when you get that speed and write the value when you have to slow, drive in the way returnig to start, reset SG when you get cruise speed and also write the mean values when you slow down. Does 40-56 (¿us mpg?) mean instant fuel consumption?, but how much fuel did you burn in how many milles?
In my car 6.2% aero improvement means 5% less fuel consumption at 75 mph. cd=0.325 to 0.30x, because crr=0.0095.
My top speed increase 2.6% (+5km/h) but mine has not the prius speed limit. But the aero top speed is relative to air in the movement axis of the car, so if you get top speed measuring movemnt relative to ground yoo need to measure top speed going back your first attempt and calculate mean of two. A wind of 10 mph will add you +10 (-7.5% cd apparent reduction in my car) in one direction and -10 in the opposite (+8.8% cd apparent reduction if tire rolling resistance neglected).
So its hard to meet accuracy on low speeds if you have low rr tires and low cd on road conditions.
Cruise control near top speed will be the best measure. (pocket Autobahn or airport needed for scientifical reasons).
I post you some semi-empirical number with my car as example:
(attached excel)