I use KProbe primarily. The main reason is that you can save the raw data in a text file and do your own analysis of the results. In fact, I've spent a lot of time in the past year working on a display tool that shows the KProbe and DVDScan results in one plot.
CD-Speed has an intermediate sampling step for the display results. The maximum count for a sample in a "display bin" (which seems to be ca. 1 MiB) is saved and shown in the plot. Erik mentioned here that the average values shown are based on this intermediate sampling. Hence the CD-Speed averages will be higher than what KProbe shows.
KProbe also has an averaging problem- the PIF averages are too high because the totals are divided by the number of PI samples (which are summed over 8 ECC blocks). This has the absurd effect that (for a bad disc) the average PIF count can be higher than the max. value.