Going from memory, I believe my CicloSport HAC4 has something like a 15 second sampling rate and 10' elevation sensitivity, whereas Debbie's Polar S720i could be user-set to something as low as 2 seconds with the same elevation threshold. We also have an Avocet Vertech which updates continuously with perhaps 5' or even 1' elevation changes being recorded.
As you'd expect, the Avocet was the most accurate, followed closely the the Polar when it was set to a 5 second sampling rate, and the HAC4 would always be 10% - 15% off what was being recorded by the Avocet. Also, because all three devices used barometric pressure, if a weather front came through during a ride that too would throw off all three units. However, even more interesting was the additional differences you'd find when you compared the HAC4's real-time watch displayed data to what you ended up with after downloading to your computer. Interestingly enough, the downloaded data was different and again, going from memory, I believe I read somewhere that the software on the computer looked at more data points that were simply stored instead of processed by the watch as a memory saving feature.
Anyway, it would be interesting to know how the sampling rates and sensitivity compare as I suspect the answer to your question may be found there.