Heres another idea which Im pretty sure would satisfy my original comment. Currently TM inserts the closing character of the pair at the same point as the user types the opening character, if on the other hand it waits for one more key press before deciding whether to insert the closing part of the pair, it could perhaps make a better guess as to whether that is what the user wants. For example; moving the cursor could be construed as a cancelling operation, so the closing character would be omitted if the user simply pops back to add a quote character to the start of a word and then clicks somewhere else in the document (or uses the arrow keys), whereas simply continuing to type after pressing the opening character would be construed as a confirmation and the closing character inserted as it is now.
so what im saying is: wait to see what the user types next, and then decide whether to insert the closing part of the pair.
does that make sense?
Chris