Ok. I did a little testing as follows
at about 8' I get random jitters of maybe 72 us in the ping time which translates to about 1" or 1%.
at 10" I get occasional 30us errors.
looking at all my data it could be that something is delaying the result of the pulsein by 25us. That something would have to be happening every 5-10 ms. I've tried bracketing the pulsein with noInterrupts()/interrupts() but the phenomenon is still there.
If you don't hear from anyone else I'd recommend getting the parallax sensor so at least between us we could answer the question.
If this is something in the avr it would be a problem with any sensor but we'd need someone with a parallax or srf to speak up.
As it stands this is not likely to stop my robot from following my cat but the sensor would not be a good choice for an electronic tape measure.
Edit: Thinking about this, the 72us is only 1/2 inch on the return trip or 1/2% and I ca't think of any mechanism for multiple 25us delays on a pulsein.
Edit again: the seeed website gives the resolution as 1 cm which would be around that 70us roundtrip variance.