Put me on ignore, then. Rest assured, I don't come here for your replies.
I'm quite aware of statistical anomalies. The problem in this case is that you've never been able to explain what is driving the anomaly, because nobody except the author - DTMAboutHeart - has access to the howevermany variables used in the regression analysis. There are also some issues with expected goals and plus minus. But for GAR, we don't know the inputs and the weighting applied to each variable. At least I don't. And so when I look at that graph, I don't just see Little as an anomaly, I see Lowry as an anomaly. Those are the most significant ones but I could probably pick apart something with every player. Aside from those two - it certainly seems to make some intuitive sense with Wheeler #1 and Thorburn last, but it's also dependent on TOI, which is a decent proxy in and of itself for who the good and bad players are (except on Maurice's teams as in the case of Lowry getting 17 mins and Dano getting zero minutes while Nic Petan gets to play with Thor).
As for Garret, I've asked him specifically about the results for Little/Lowry before and he thought it was strange as well. So no, I haven't fully bought into WAR/GAR as a method of comparing players. I respect the work but I'm not going to become a believer until it all the variables and weighting scan be explained. I don't know what percentage of the advanced stats community believe that GAR is better than Corsi, but I don't think it's the majority.
No, you inferred that since I only use Corsi, "I must think X". I don't use only Corsi. I believe I was the first one to point out that Ben Chiarot was junk back when his Corsi numbers were excellent. I've been wrong about several players and I will be wrong lots in the future, but on Lowry, I don't think I am.