I have found blocking the grille helps a LOT - up to 2 mpg in the case of my cars.
MetroMPG is the guru of testing, but for coastdown testing, I always try to minimize all variables other than what I'm testing. So for aerodynamic drag, pump the tires up as high as you dare, so rolling resistance is minimized, and remove all extra weight possible to remove as much "momentum factor" as possible (no "assistant" riding along). I also like to rig it for a longer duration so errors starting & stopping the stopwatch and reading the speedometer are less of an influence.
Then I try to MAXIMIZE the factor I
AM testing. So since aerodynamic drag increases so much with speed, start the test at as high a speed as you dare. Also, for the same reason, try to end the test while you're still going at a fairly decent speed. This probably wouldn't be workable for you, but when I was testing aerodynamics on my salt flats race car, I found a desolate stretch of highway and tested from 110 mph down to 60 mph. I was fairly sure aerodynamics would be significant at these speeds, and it would take between 50 seconds and a minute to coast down those 50 mph). I also had my tires pumped up to 70 psi and tried to compensate for the gas I burned by putting a gallon in every couple sets of runs (with the gas can left alongside the road during the runs). Of course, that car probably burned gas at a much faster rate than your eco-modder car.
I had a GPS speedometer and took it a few mph above my "start speed", then threw it into neutral & let it coast until down to the target speed when I'd hit the stopwatch. If your regular speedo is steady, that should work ok, but mine was really bouncy. I also set out orange cones to mark the "start point" and "end point" a mile apart. The tests are done in both directions so wind and grade will cancel-out. I'd go a little over a mile past one end, then turn around & accelerate up to start speed (around 112 - 114 in my case) & try to time it so I was down to 110 at about the start cone. Then do the same thing going the other direction.
I like three "sets" of two way runs for a good average. Start without the modification, do three sets, then attach the mod & do three sets that way. Then finally, take the mod off & do at least one more set to confirm you're still getting similar results you did in the first group. That's the "A-B-A" technique. The first "A" is without the mod, the "B" is with it, then the second "A" is without the mod again.
To compensate for the weight of the mod (since extra weight helps it coast longer and could affect accuracy), I'd either carry the mod inside the car when it wasn't attached (if it would fit), or else carry an equal amount of ballast weight in the car if the mod was off and alongside the road with the gas can.
I was pleasantly surprised the results seemed to end up really good. I could definitely see enough consistency to tell which mods worked best (longer coastdown times mean less drag, right?)