My only "source" is the scans I perform myself, which is the most trusted source available to me.
When performing DVD Disc Quality Scans with a LiteOn drive, not all possible samples are actually read by the drive and passed to the software. The exact reason for this is unknown to me, but the number of dropped samples increase with a higher scanning speed, and it also increases when using slower connections such as USB and FireWire.
The number of dropped samples can be seen by calculating the total number of possible samples and comparing to the samples reported by CDSpeed. The reported sampling interval at the end of scan, e.g. 1.33 ECC also indicates how many samples were dropped.
Example: The drive/interface/software drops 1 out of every 4 samples, thus actually seeing only 3/4 of the samples. The inverse of 3/4 is 4/3 = 1.33 (approximately) and this would be the reported sampling interval.
There is some mechanism in CDSpeed that calculates PIE and PIF averages by taking the actual sampling rate into account AFAIK, but the totals counts will be lower when the number of dropped samples is higher.
I can get sampling intervals as high as 2.40 ECC when scanning at 16x with an external drive, which means that only 42% of all ECC blocks were sampled and the remaining 58% were dropped.