Here's how I read them:
The top "PI Errors" graph (PIE for short) represents all of the DVD data errors as they are read from the DVD. These errors are shown before the first line of defense DVD "error correction" data has been applied to fix them. All DVD rom drives internally apply error correction information automatically, so Windows doesn't know or care that it is happening.
The bottom graph, the "PI Failures" graph, represents any errors that remain after the first set of error correction data has been applied. Since the "first line of defense" error correction logic failed to correct all of the errors, the remaining errors are called "PI Failures" or just PIFs for short.
Fortunately, there is another layer of DVD error correction data burned on the DVD (it can be thought of as a second line of defense) that will be applied against the remaining PIFs. As before, this correction logic is internal to the DVD player itself - Windows doesn't know or care that it is happening. This second layer of error correction is all that remains between a good burn and a coaster, so it must be 100% successful. Any failures in here will count up the "PO Failures" box located near the bottom-right. Note that "PO Failures" are only counted and not graphed like the others, since "PO Failures" are not allowed for an acceptable burn.
How I interpret the graphs:
This interpretation would apply for a graph scanned on the BenQ 1620 at 8X (or maximum, which is the same as 8X). You can also scan at 12X and 16X, but the errors would rise significantly, particularly for cheaper discs, so you would need to keep this in mind when interpreting the results.
I like to look at the graph to see how even it is - particularly the top "PI Errors" graph. Problems usually occur near the end of the burn, so I carefully look at the end of the graph to make sure there's no error peaks. It's very important to read the graph scale along the left. I consider a burn to be very good to excellent if the PIE maximum peaks below 20 (your burn peaked to 9, which I consider to be excellent). Even then, I don't get too excited as long as the maximum PIE peak is below 60. I've seen many cheap discs burn with PIE peaks of 200 - 300, and many of them read back fine.
But by far the most important graph is the bottom PIF graph. For a 1620, it's important that the PIF maximum peaks do not exceed 16 or there is danger of a coaster. It's important that they do not exceed a maximum peak of 32 or a coaster becomes more likely. The current version of CDSpeed looks only at the PIF maximum when it calculates the "Quality Score". For example, your PIFs peaked to a maximum value of 6 which equates directly to an excellent CDSpeed "Quality Score" of 97%.
Other rule-of-thumb items which are not as critical:
a) I like the PIE totals to be below 100,000 (yours are excellent at 8817)
b) I like the PIF totals to be below 2000 (yours are excellent at 569). I've seen them as low as 21 for some of the best discs - but not often.
c) I don't like to look at the PIF graph and see what I call "clustering". A cluster would be a series of repetitive PIFs that cause the graph to be coloured solidly with PIFs for about 1/4 inch or so. Clustering often occurs near the end of a burn, and can cause reading issues on some players, particularly if the PIF cluster maximum peak is above 10.
d) I like to see even jitter levels across the whole graph averaging around 8.5% and peaking below 14%. Often, jitter will tend to rise near the end of the burn.
This represents what I consider to be a good burn. Others may well have different opinions.