I agree that driving as straight as possible would certainly remove some of the variables from the test. According to Garmin's site, the sampling rate of my unit is once per second which seems like it's not going to be clipping off big chunks of my route - certainly not to the tune of half a mile after 9 quarter turns (which is how many I made on the test). I would guess less than 150 ft of error from the turns, if that much.
Anahymbrid is dead right about the over-inflation and I stand corrected. Larger circumference = you've gone farther than than your odo reads which wasn't what I saw. So I agree over tire inflation is not the culprit and, if anything, it might make the error less severe.
So another theory, and a weak one, is that tire wear (only 5000 miles for me) has decreased the circumference.
A more likely theory, and I've heard car makers do this on purpose, is the speedometer just is calibrated to read slightly high. Honda wouldn't want to get sued by people who got tickets for exceeding 65
MPH when their speedo was indicating 65
MPH. So to protect themselves (and us), they make the speedometers (and accordingly, the odometers) read higher than actual.
It would be great to get a few more data points from other HCH owners to see if this affects the larger HCH population.
Maybe the good news about this is my 36,000 mile warranty should be extended by another 809 miles for me (and probably most of you). You just need to prove it when push-comes-to-shove.
cheers,
-dan