#1 and #2 are great answers that have been shown by testing to be true.
The only test results that have been presented in the past 5 years (current crop of GPS units), that I am aware of, have shown that when there are no tall buildings near the course the distance measurements reported by GPS units are no more than 0.5% in error, and usually a good bit less.
It has been shown in two sets of tests carried out by two different groups that people in races run 1-2% longer than the actual SPR distance of the course. That's really all that needs to be said. And it makes everyone happy because it means that the person reporting his GPS measurement probably DID run 1-2% longer than the stated course distance.