I don’t think that the presence of tags you don’t use “pollutes” the XML.
Indeed, my understanding of Tim Bray’s approach to defining XML in the first place was that XML processors need not handle tags with which they are not concerned; XML provides a straightforward and disciplined way to delimit information with which some processors will be uninterested, and lets them ignore that information without needing to parse that information or know anything at all about it.
If you like, you can use XSLT or any other XML processor to remove unwanted RTF tags. I think Tinderbox will handle this gracefully, though I’m sure that’s untested. As you will read in The Tinderbox Way, however, the decreasing cost of storage suggests that the time you spend doing this may not be theist investment.