The problem is the argument of how valuable context can be. WAR is great in that it eliminates context from the equation, just tells you how valuable the pitcher was, whether it was pitching 5-8 innings at a time, mopping up the mess from other starters or closing out critical games/pitching in almost exclusively high leverage. WPA and other similar context-based stats will tell you how much value the pitcher contributed when put into different indexes of leverage. This is where closers and similarly elite relievers will get a huge boost, because they usually pitch in the more valuable, game-changing high leverage situations.
Now, I don't know if any one of them is better than the other, a combination of both will probably be the optimal use here. However, something to consider is that a top tier reliever's WPA will basically be artificially boosted because they are generally inserted into the most important situations in a game. They got through that inning or that batter, which is great, but who's to say that a Joe Biagini couldn't have done that job more times than not rather than pitching the lower leverage mop up/trying to keep the lead/deficit to a max/min? This is how a guy like Zach Britton can have a higher WPA than every single American League pitcher, and every Nationa League pitcher, including Clayton Kershaw and other elite starters. Shouldn't that point to a flaw in context based stats? I think that's where we have to consider how much to take from each.