It really doesn't matter how many tags are assigned css classes (even if it's every single one) in the source that's going to be used for building a mobi. Because all the css gets converted to simple HTML 3.2 attributes in the mobi (all the attributes that are supported by mobi anyway).
Before you're convinced that it's necessary to strip all of those classes from your source epub, use mobiunpack.py to see what's inside a normal mobi... all the css classes and a css file would be far prettier than what's actually in there, I promise you.