This was addressed to Taco but I figured I'd weigh in case my thoughts might help any. I've done a fair bit of work with comparing goalies to backups, you can search my blog if you want to find some of them, but those studies run into two main difficulties: Small sample sizes for backup goalies playing with workhorse starters and difficulty in evaluating how good the backups are, particularly if they had short careers or didn't play a lot of games for other teams. It's possible to adjust for the latter issue (e.g. look at what the backup goalies did on other teams or adjust backup goalie stats based on their career averages), but the sample size thing is tough to account for and really restricts the validity of backup comparisons for goalies who played nearly all their team's games.
It does work fine for cases like Smith/Resch or Fuhr/Moog, two goalies who posted pretty similar numbers while platooning on the same team, although the analytical problem there then becomes trying to figure out whether the more highly regarded goalie was perhaps overrated or whether the less heralded one might have been better than commonly thought, or if there may be some other factors clouding the issue.
I think it is mostly a good assumption to rank a goalie on a weaker team higher than a goalie with stronger teammates if they are rated similarly or have similar numbers, because for most of NHL history the stronger teams tended to allow fewer and easier shots against and typically the voter bias come awards time was in favour of goalies on strong teams (GAA was clearly the main driver of First Team All-Star voting for much of league history, for example). I also tend to think that playoff team success typically has too much of an impact on a goalie's reputation in general. That said, you can't completely discount goalies posting great numbers on great defensive teams either, as I mentioned in the Thomas discussion upthread.
I'm not aware of any public site that has all the unofficial stats in one place, but complete goalie stats from 1952-53 to 1982-83 were compiled by Roger Brewer using the data from the Hockey Summary Project and the Hockey Databank. He posted the file at the
Hockey Analysis Group in Yahoo Groups, you could either join that group or email me and I can send you a copy of the file if you want.