Open a copy of the TMX file in TextWrangler (a UNIX app if the file is larger than 380 MB), and use grep to delete end punctuation. Use CT's TM editor to get rid of duplicates.
Alternatively (easier), convert the TM to plain text. Open in TextWrangler (a UNIX app if the file is still larger than 380 MB). Use grep to delete end punctuation. In TW, use the built-in duplicate remover. In CT, import in an empty TM.
Not sure if this is cool enough, though... Let's ask that Michael if that's the case. He's the cool expert.