This season, in the games Martin Brodeur has "won", he has given up 5, 4, 1, and 2 goals. In the games that he has "lost", he gave up 2, 3, 3, 4, 2, 3, and 1. Which leads to this unusual split:
In wins: 3.00 GAA, .876 save %
In losses: 2.56 GAA, .906 save %
The other 18 skaters on the team have a very large impact on whether the team wins or loses. Usually much more so than the goaltender. This has certainly been the case this year in New Jersey, and the reason has nothing to do with goal prevention but rather what is going on at the other end of the ice:
Brodeur's goal support in wins: 5.00 goals per game
Brodeur's goal support in losses: 0.71 goals per game
Goalie wins ignore half of the inputs that determine wins and losses (goals for), and leave the other half (goals against) unadjusted for number of shots against and shot quality. With all that noise clouding the data, it is impossible to tell from his win totals if a goalie is very good or just playing for a high-scoring or great defensive team (or even both). Brodeur's season to this point is a great example that goalie wins are not very meaningful at all in terms of evaluating goalies.