lol what? You think people are manually assigning xG values to every single scoring chance in every game? Are you serious?
Shot locations get input by someone, yes, and sometimes have errors. Each website with a stat has different methods of computing xG but they pretty much all use the same shot locations provided by the league and assign a goal probability from the location. Some models - not sure about public models - supposedly use other variables like time since the last shot to identify rebounds, and in principle you can get skater speed and location and track passes and so on, but I don't think the public websites do that.
And anyway, all of that is mostly irrelevant. Benoit has a 42.5% CF% and 44.6% FF% 5v5, by far the worst of any regular D. His actual GF% is 52.6% because he has a very high on-ice sv% of 93.8%, second only to Timmins (who he plays with more often than not). Actually you can look at stats for
defensive pairs and see that Benoit has poor possession stats with everyone he plays for save for 30 minutes with Myers. If you only care about goals then I suppose you can conclude that the Benoit-Timmins pairing should be broken up because they both do better with other partners (the difference being that Timmins has even better CF% with other partners while Benoit has much worse).
As for context, I don't know how that helps. Benoit doesn't get the toughest matchups, McCabe & Tanev do. It's also not inevitable that defensive defensemen have poor possession stats - Tanev's are fine and McCabe's are the best on the team. Of course they're both much better players than Benoit, so no surprise, but it's possible for 3rd pair D to break even. And yeah, Benoit gets more D-zone starts than not because... he's not any good in the o-zone. The comment about PP time is irrelevant; we're only talking 5v5 here.