Are there though? Is the proportion of people who are likely to vote but undecided really a substantial category compared to the proportion of people who know who they'd vote for but don't bother to do so?

I'm not saying anything about that comparison. I'm just saying that, yes, people who change their votes do exist, and they exist in numbers large enough to matter. (See, for instance: Republicans who voted for Biden in Georgia and Arizona more than made up for the ~10k vote margins.)