Probably a valid theory, but I misspoke, I ADDED 2% to the comp distance. The GPS actually indicated a farther distance than the bike comp did, so if this did come into play, I should have added even more.

My bike comp reads to within less than a tenth what any of our three cars do for the same route, so I'm OK with it anyway.

I think riding a known distance might be more accurate than measuring rollout; tires compress and squirm a bit on pavement under load. It's probably not a lot, less than a mm or two per rev, so probably not worth worrying about.