I'm sure Keson's accuracy number is for an uncalibrated measurement. That's the number most people buying a measuring wheel would want to know.
Both Pete and I have measured tracks with a calibrated measuring wheel and for both of us our two measurements were well within 0.08%, and the measurements were very close to what they should have been for a 400-meter track. When it is very easy to follow the SPR, like on a track, I think you can get very accurate measurements with a calibrated wheel.
We assume (for good reason I think) that it is much more difficult to follow the SPR on a road course with a measuring wheel than with a bike. But I don't think there is a lot of experimental evidence to confirm this.