Looking at that list, I'd say it's a pretty flawed methodology. Nate Robinson + four average players has a 53.1% winning percentage? Perk + four average guys wins 53.5% of the time? I don't think so; those sound a lot like lottery teams.
Player win % is an attempt to combine Offensive Rating and Defensive Rating into a single statistic. I believe it should be taken to assume that an average player would have a .500 player win %. Keep in mind that a 44-40 record, or two games above .500 is a 53.7% winning percentage. I have no problem believing that Nate Robinson is better than the average production that teams get from their point guards.
These are also the un-updated numbers. The formula was tweeked because it was believed to undervalue three-point shooting. Thus, players like Nate Robinson and Ray Allen have a higher value than I listed, while players who lack an outside shot, like Tony Allen, go down in value.
My main point was that advanced metrics suggest that Tony Allen is better than other wings described as low-offense defensive specialists and a reasonable interpretation is that he actually is a better defender than those players. You could make a reasonable case that Tony Allen's defensive abilities are better than Dwayne Wade and on par with those of Gerald Wallace. The problem is that his offensive abilities make it often look like an average player being guarded by Gerald Wallace. He wasn't worthless, but he had a skill set which made him incredibly useful in some situations and worthless in others. He's more useful than a 3pt sharpshooter who can't play a lick of defense.
I think Tony Allen wanted to be the first wing off the bench no matter the circumstances and to be automatically inserted into that role whenever he comes back from injury. I think there is a risk that he would have turned into a malcontent trying to prove he deserved that role, leading him to try too hard on offense or injure himself, if he had stayed in Boston.