That's a fairly bizarre choice, given that Libertarians in general have been polling unusually well this cycle, and given that recent SurveyUSA and PPP polls have given Paterson 5 and 7 percent of the vote, respectively. Yes, Republicans are trying to knock Patterson off the ballot, but while Libertarians aren't particularly good at actually winning any elections, they usually are quite adept at securing ballot access for themselves.
And yes, Patterson is likely to wind up with a smaller share of the vote than he's getting now in other polls, but he can certainly have an impact on the race nevertheless, especially since he's liable to scoop up some anti-McConnell protest voters who would never pull the lever for a Democrat.
So CNN's poll simply isn't reflective of the choices Kentucky voters are likely to have before them in November, which means they aren't polling an actual election—just some hypothetical contest that doesn't exist. And in any event, there's no reason not to ask two versions of the horserace, one with Patterson and one without.
Some analysts fetishize independent polls over partisan ones, but what matters is the overall quality of the survey, an assessment that involves many criteria. And sometimes, it's as simple as making sure you ask the right questions. CNN didn't do that here, offering a good example of why it makes no sense to prize non-partisanship above all else.