In 92, while Perot brought in some Dems from Clinton and many people who probably wouldn't have voted otherwise, it's hard to argue he didn't take a lot of votes away from Bush I, maybe the sum total once you remove people who wouldn't have voted for anyone, and account for those who would have voted for Clinton, was enough to have changed the outcome on election day in 92 but in 96? Not very likely.

Now one can argue that Perot's visibility and all combined factors drained attention and thus votes from Bush I and Dole but I think that has to be considered a different thing all together. The first concept says 'had this man been removed from the ballot, as in a runoff, how would the vote have turned out?', the second essentially posits removing them from the race entirely and you could apply the same argument to hotly contested primaries. e.g. had Gingrich/Perry/Santorum dropped earlier or had Ron Paul endorsed, would Romney have faired better? And while the answer is probably yes it is a different kind of thing.