That's one of the inherent limitations of save percentage. (To be clear, in general, I think it has a fair amount of informational value - but this is a good example of its limitations).
In game 7 against Toronto, Joseph allowed 4 goals in the first period (included a back-breaking goal with 20 seconds left). I'd have to go back and watch the highlights, but from what I recall, Joseph didn't look great. The Blues were demoralized, and the rest of the game was lifeless (Toronto scored two more goals in the second).
It's true that Joseph's save percentage took a hit because of this game, but it doesn't capture the timing of the goals. Regardless of what else a goalie did, allowing four goals in the first period of a game 7 is essentially a series-ending performance.
Roy's save percentage was a bit lower (but still very high for the era). There's some evidence that Roy lightened up in games that were out of reach. (For example, on May 24th, he allowed two goals after the Habs were already up 5-0 late in the second period. Or on May 16th, he lost his shutout with just over a minute remaining, with Montreal sitting on a 4-0 lead). And, as we all know, this was the year that Roy helped his team win 10 consecutive overtime games.
In 1993, the Habs were a better team than the Blues, and we can't give the goalies sole responsibility for the outcomes. But looking just at an averaging statistic (save percentage) misses the point. Joseph had his worst game of the playoffs at the worst possible time, and Roy stepped up when it mattered, and took his foot off the breaks when it didn't. That's important context, and save percentage doesn't tell us anything about that.