Meh. Without sounding awful, Ottawa is regularly first in PB's chart but in the two votes I've seen has finished last in their division.
Assuming these numbers are even correct and measuring what they're supposed to (which is a MASSIVE assumption), there are so many things that they don't capture. Outside of maybe baseball, there are just far too many elements and moving parts to sports for even a giant spreadsheet to capture things adequately.
If you build off of a set of any stats, you're probably not going to end up with a good team plain and simple. That's why scouts still have jobs. Intuition and our eyes are still generally the best tests for player abilities.
1) Data taken from three seasons against quantifiable "elite" production using a combination of actual production, GF% and two other metrics weighted to their correlation to GF%. It's not perfect, but it does describe how well a player tends to do in terms of possession, generating dangerous chances, and potting points.
So no, it's not perfect, but it's a decent guide for which way the ice tilts when a player is on.
2) I only used it as a reference for my picks. Every single line and pairing that I have stylistically work with players occupying complementary roles in their actual, real positions.
People keep missing the second point. For example, Kreider-Danault-Nichushkin is a line with a monster net crasher, a centre with great support play and positioning, and a guy who's great at both transporting the puck and forechecking.