1. You want to evaluate goalies SV% at even strength. Their overall SV% gets thrown off if they face a disproportionately low or high amount of shorthanded situations. Overall SV% last year was .92282 at even strength.
2. Even in one season it's not a big enough sample size to get at the goalie's true level. You see a ton of variance year-to-year - the difference between a Vezina performance and an average starter is a goal every 8-10 games or so, which can have a good amount of luck in it. Fleury's even-strength SV% the last few years: .9193, .9296, .9147, .9267. Most goalies who are not Rask or Lundquist have similar variance just due to sample sizes. So MAF is probably a little better than average overall.
3. Fleury does play a lot of games, so you can spend less on your backup and if he is bad it's not going to hurt you as much.
4. He has had some embarrassing playoffs, but those are a small sample of games overall. If you believe in some kind of clutch thing you're going to mark him down. Of course, you may also think his problems are fixed now. Overall I think he gets a slight knock here, but not very much.
Overall I think he is league average, but can basically be depended on to be that with very little risk of collapse.
So the question becomes, can you find that league average production elsewhere and how much would it cost. There have been a lot of goalies who have come out of nowhere recently, but does Pittsburgh have one of those? Can they identify him and acquire him cheaply? For a team that is basically a contender every year so long as they don't put complete shit around Crosby and Malkin, I think you want to be risk-averse and pay for the security of having league-average goaltending even if it hurts you a bit elsewhere. So, it seems about right.