I don't think that nominating more liberal/conservative candidates in hostile territory would have all that big a change on elections. It might help slightly at the margins, yes. But the politics is about more then triangulation. As an example, I'd offer my home district of MD-1st, which . That means that if the nation were exactly balanced between liberals and conservatives in a given election, you'd expect the republicans to win this district 63-37, i.e. a landslide.
Despite being in this conservative hellhole however, a democrat managed to win in 2008 but he didn't do it with a really conservative campaign. He was pro-choice, pro cap and trade, pro stimulus, voted for one of the two healthcare bills out of the house, supported the financial regulations and voted for hate crime legislation to protect gays. Yes, he tacked sharply right on guns, stimulus and a few issues of local importance (fairly conservative stance on the 2010 Chesepeake bay dredge report) but his liberal positions far outweighed his conservative ones. But despite having a pretty liberal track record, he outperformed the PVI handicap by 10 points in both 2008 and 2010.
So the reason why parties don't always triangulate for the center isn't just to appease their base. There are a lot of things that make a candidate electable or unelectable beyond triangulation. So it makes sense to pick a guy who would actually advance your agenda and is slightly less likely to win then to pick a guy who is still a real underdog but wont be any good even if he does win.