I know what "regress" means (as my next sentence made obvious), but what are you claiming? That it's impossible for the electorate to shrink in a way that decreases the proportion of non-white demographics more than their 18+ growth rate?
With that said, unskewing was totally based on the concept of party ID and nothing else. If you are talking about the actual demographics of the election, then that is where polls can end up being wrong. I'm talking about what Gallup did when they assumed that the 2012 electorate would look like 2000. It has nothing to do with party and everything to do with demographics.
Unskewing was not just about party ID -- quite a lot of it also involved insisting that, regardless of what Likely Voter polls showed, the electorate would revert to have many fewer black voters (and non-whites in general) because 2008 was an unusually elevated turnout. You're also doing a touch of that. It's reasonable to weigh by demographics, but it's not reasonable to assert with certitude that the electorate will look like 2012 even if, say, polls are showing depressed enthusiasm and higher rates of failing LV screens among certain groups.