But isn't it a little bit odd that pollsters refuse to include Gary Johnson in their polls when he's actually on the ballot in NC ? A previous Zogby poll has shown that Johnson could take away about 2-3% from Romney in the state.
No, it's not odd when you consider the historical pattern of Libertarian polling. They always vastly overpoll, with the margin narrowing as Election Day nears, and on Election Day they get considerably less than the margin of error that the polls have. Most third party responses in polls come from people who don't like either major party choice and thus the response should be interpreted in that fashion. By and large, those people either vote for one of the two major parties come Election Day or they stay home.
So by asking a longer question that slows down the poll and increases the chance of being hung up on, all a pollster gets is statistical noise. For a few states such as New Mexico (Johnson) and Virginia (Goode) it is worth asking about them, if only to check if they actually have a shot at doing better than the MoE.