All these comments and tests have made it clear that when a runner uses a GPS they will not give accurate results.
Bob Baumel has pointed out the following on a GPS:
A runner that uses a GPS can have errors on the order of several meters (sometimes as much as 20 m)—and that’s just in measuring the location of a single point. Errors dealing with course length are much greater, as the course must be approximated by making point measurements at some number of points along the course, and the overall error depends on how densely the points are sampled, the errors in locating the individual points (inevitably mis-positioned somewhat to left or right of the true path), and the algorithm used in calculating the course length from all the point locations. Also, some points along the course may not have clear views of the satellites used in making GPS measurements, resulting in very large errors in locating those points.
The RRTC/USATF as stated by one of the responders needs to have something posted on the fact that a GPS in most cases will produce long a measurements. We are currently working on coming up with just that.