Which would also be safer, most likely, since you wouldn't necessarily have to pay attention to your speedo as much as the road and things going on around you while you're speeding beyond the territory of instant jail time at your "top speed" for aero testing
I really don't think the ocifer is going to like it when you tell him that you're just performing a simple aerodynamic test at 120 MPH on a public road. Just me though - maybe you can get away w/ it LOL.
I would think that testing up to 80-85 MPH should cover just about any possible daily-attained situation. Anything beyond that is a moot data point that is only considered as a non-skew factor for the rest of the data, to show linearity.
Will you be taking readings on how long it takes to drop per increment of speed, or how many MPH dropped in increments of time? (Check speed every 5 seconds or check time every 5mph?)
It might be interesting to make both graphs and compare the plot point densities of each graph. They most certainly won't line up between each other, but it would help create an "average" line between the two graphs, which should be more linear than either of the initial tests.
This is where having a video camera comes in handy - you can draw up both data points while only having made one run. Don't forget though, you'll want to make this run 3-4 times, to properly report the results, then average all the runs into the same plot of data points, creating an even more linear average.
/ramble
OK, I'm done LOL.
Seriously though - if you're going to do something like this - keep it safe.