Being involved in a business that does GPS mapping/survey that guarantees a data accuracy less than 4cm and aims for 2cm whilst moving, I have a little understanding of this issue.
First of not all GPS's are created equal. The gear we use, uses the Russian Glonass as well as U.S. satellite system and is also conected by radio to a base station to give a real-time correction. We consistently achieve 2 cm accuracy. Hand helds, car systems and basic aircraft GPS all vary from 1m to 35m accuracy and worse in some cases.
The point he Mark1234 raises is very valid though. It is more about 'repeatability' rather than variability. Even the 35m accurate systems will probably get you back to the same spot, pending clarity/accuracy of the screen read-out, within a short time period. 2 hours later, next day, change in the weather, satelites in view and position can very quickly see that position change however. What they will not do is reliably take you to a pre-detirmined spot on the earth with any accuracy. Thus the position of a 10m accurate system can vary 10m radius from the given point or up to 20m from a point marked previously.
Given all that I would generally trust the GPS more in a steady state on level ground. Tyre pressure and wear does have an impact on the speedo reading in my car as well. As they wear the difference tends to go from an indicated 2kpk faster than the GPS to 4kph when needing replacement. If you were on a 45degree incline then the GPS would obviously be under-reading drastically.
If you wish, it is easy to use a stopwatch to check your speedo. I once had a car that when I first drove it I seemed to be passing cars all the time even when driving at the limit. A police radar fine a few months later alerted me to the fact it under-read by 12 kph. This was traced back to the wrong drive gear fitted to the speedo gearbox drive. Still had to pay the fine though.
My thought anyway.
