That’s an interesting math problem. Assuming a GPS deviation of up to 10′ (which seems reasonable), your deviation for a 1 mile trip could be as much as 20′. Since there are 5,280 feet in a standard mile, that’s a possible error of as much as 0.379% (as in about one third of 1%).
That same 20′ max deviation for my 42 mile commute results in a max error of 0.009% (or less than one one-thousandth of a percent).
Technically speaking, I’d wager that we’d find more deviation error and variation based between folks you ride on the outside of turns vs inside over the course of my commute. But that would be getting pretty fussy