Yeah, I was using the single sample read as well but decided to try this after you mentioned reading 100 samples every 10 msec. Reading 3 samples produces somewhat the same result as the single sample read in that there can be 2 or 3 or 4 count differences between changes in count value. I don't think pulses are skipped but rather the 'implicitly timed read' you referred to earlier. I'm going to experiment a little with the pulse rate and see what effect it has on the reported values.
