The question clearly isn't who they prefer, but "who did the debate make you more likely to vote for?"
Or, since you can't see it clearly - "did the debate change your mind about who you were already supporting and in which direction?"
Certainly, anyone who supported neither candidate, would be in that "neither" category. But the question doesn't count them at all as separate from people for whom the debate didn't change their minds.
This question doesn't measure who supports who, it measures who switched sides.
18% of those either undecided or supporting Romney (not likely) now lean Obama.
35% of those undecided or supporting Obama now support Romney.
47% of those undecided or already supporting either Romney or Obama were not swayed either direction by the debate - meaning, they are still undecided or they remain locked in to their choice.