better UCD treatment of characters not allocated by Unicode