One thing I forgot to mention. CDSpeed provides a "Quality Score" number located near the lower right-hand side of the "Disc Quality" window. This quality score is based totally on the PIF's and completely ignores the PIE's.
For example, on the BenQ 1620 if you have a PIF spike that hits a maximum of 8, your quality score would drop to 95%. On the LiteOn 451, if you have a PIF spike of 2, your quality score would similarly drop to 95%.
Basing the CDSpeed quality score totally on the PIF's (and not at all on the PIE's) is somewhat controversial, but there is some reason to it. All DVD's write extra error correction data that's used later on when the DVD is read to try and correct these errors. This is fortunate, since the PIE's represent these errors as they are read off the disc, and there are typically a lot of them - even with the best discs available.
Once this error correction data has been applied to correct these PIE's, if there are still any errors left they are shown as PIF's. Hence the term "PI Failures" - meaning that this "first line error of defense" error correction logic has "failed" to correct some of the "PI Errors".
The remaining PIF's are then fed to a secondary (and final) error correction unit, and yet additional error correction data is applied. If the PIF's aren't too bad, then they are all corrected and you have a good disc.
If the PIF's cannot ALL be corrected, then your "PO Failures" (middle, bottom) count box will begin to count up, and you will end up with what might be a "coaster" (although a slower read speed might still allow the disc to be read).
Note that unlike the "PI Failures", the "PO Failures" are NOT graphed by CDSpeed, since they are usually fatal - they are just counted instead. Their count should be 0 for a good burn.
I've attached a scan of a Khypermedia CMC Mag F01 disc originally written on a NEC 3500 as an interesting example. The PIE peak near the end of the graph is only 55, so if you were only to look at the PIE's (or if your burner could NOT report PIF's, like the BenQ 1620 with an earlier firmware from only a few months ago), you might think that this burn was a good one.
However these PIE's are NOT of the correctable variety, even though the peaks are well below the so-called "PIE below 280 acceptable standard". Besides the obvious "PI Failures" problem, notice the "PO Failures" count of 5818.
Incidentally, if I'd written this disc on the 1620 instead of the NEC 3500, it would have read back fine. Also if I'd written an "HP" CMC Mag F01 disc (same Media ID) on either the NEC or the BenQ at 8X it would have produced a near perfect burn.
Hope some of this is useful!