RRH is crap. During the Wisconsin recall, they put out a bogus poll that showed the Democrat leading by only two points and he went on to win by ten. The narrative that they are trying to create here is that all the undecideds will break for Vitter and he'll win in the end, but this is really weak since it barely keep Edwards under 50%. I'm not surprised they'd stoop low for Vitter, but they didn't stoop low enough to get him above a miserable 42%.
Can you provide a link to the Wisconsin poll? According to RRH themselves, the first poll they ever conducted was of the runoff in LA-3 in December 2012 between Charles Boustany and Jeff Landry, which showed Boustany leading 51-33 (he went on to win by 61-39) (link here:
http://rrhelections.com/index.php/2012/12/page/7/ ). Indeed, the only incorrect winner they have ever projected was in a poll conducted shortly before the 2014 elections in Arizona, which showed Democrat Ron Barber leading Republican Martha McSally (of course, Barber went on to a defeat). I'm not doubting you; I just can't find any record of the poll you're referring to existing. (Also, you might've forgotten, but Republicans won every statewide recall election in Wisconsin during the recent hullaballoo, being defeated only in more local state Senate races).
