Again, as a quick and dirty metric save % is fine. But a simple thought experiment shows its flaws: our opponent has only 3 shots on target, all of them from close range and all of them rifled into the top corner. Allison is perfectly positioned, anticipates each shot perfectly, and covers as much ground as humanly possible but the shots are hit with too much velocity and accuracy. Save % = 0
Conversely, our opponent has 10 shots on target. All 10 are hit from outside the box, tamely, and straight at Allison, who loses concentration and lets one of them in. Save % = 90.
Everyone can see the “first” Allison played better, but save % indicates exactly the opposite. Now over time most of that “noise” gets filtered out with a large enough sample, but there is still noise. Edwards and co. will have something like save %, but without the “noise”.
Why use cluster bombs when you can use a laser-guided missile? 
If he's perfectly positioned in the first example, then how did they score?
Goalkeeping is positioning first, anticipation second, footwork third, and hands as a last resort
If the first three shots are from close range, and Becker is "perfectly positioned", then those shots don't score, because a perfectly positioned keeper in that instance is right on the toes of the attacker, and probably taking the ball off the attackers feet as soon as they make their backswing.
I know where you're coming from, but the realities of the game have a way of showing up fancy statistics. There are fundamentals that are just necessary at all levels of the game, and a keeper who concedes shots on target from any distance, angle or power, at less than 66%, is a keeper who has no real utility for the team. So while I do agree we can't just use save percentage alone, we also can't dismiss it as "not being advanced enough", because it does tell a story about a keeper, and can give cause for optimism or cause for concern, depending on the number. There's a historically robust standard that we know about save percentage, and we can at least apply it to, on one hand filter out noise (sub-standard keepers, for example), and on the other hand, look for consistency of performance.
Having said all of that, we could go around and around talking about the merits of save percentage, and it's an interesting topic in the "stats vs the game" conversation, but we got here from the debate about whether elite performers are able to sustain "exceptional" levels of performance for more than one season. If so, then we can't be worrying about regression to the mean. If not, though, then we probably should be thinking about replacing the key players every two seasons, no?
