Just wanted to chirp in and say that GF% vs xGF% is a pet peeve of mine. I don't think either of those stats actually tell the story.
As a coach, I might recognize that my player got lucky by giving away a high-danger chance and being bailed out by his goalie. He's still sitting on the bench even if it didn't show up on his GF%.
On the flip side... if I have Shawn Horcoff or Marty Reasoner getting grade-A chance after grade-A chance in a prime scoring position... it's still not going to make me feel any better when their xGF% stats are looking better while they muffin yet another shot into the goalie's crest.
It's a double standard, but on offense I want results ("bear down... find a way"). On defense I want to minimize risk ("smart decisions").... but it speaks to my point that neither of these stats are all that useful, but if I have to choose, I'm looking for results rather than expected results.
I think in both cases both numbers are valuable.
On offense, you have some guys who for the life of them can't cash but generate opportunities left and right. You point to Horcoff and Reasoner, but the guy that comes to my mind almost instantaneously is Josh Anderson, Paul Byron, Patrick Marleau or guys like Glenncross and Mangiapane. These players have no real finish, but are clearly guys that generate possession and chances; if you put them beside the right player (say a Tyler Toffoli or Johnny Gaudreau) Then suddenly they move the dial for you quite nicely.
On the other hand, you have guys that are clearly elite level finishers that could cash on anything semblancing a scoring chance, but are rarely the kind of players that create these opportunities themselves. Guys like Patrik Laine, Tyler Toffoli, Jonathan Cheechoo or Steven Stamkos.
Both of those niches are extremely valuable if you lack the other one.
As for the defensive side of things, I think there is something to be said for not actually getting scored on. Obviously if you just have high possession and don't let scoring chances through then you're great defensively.
But I think people sleep on players that consistently outperform their xGA in their actual GA. A guy like Russell is the perfect example of that. His possession is awful. But he's made a career of being trusted to not let the puck in his net, despite having bad possession and giving up a similar number of chances to his contemporaries. But... he still just manages to not get scored on despite this and has for almost 1000 gp over the course of 16 years. Clearly it's not luck. And I think that has a lot of value.