closingpunct() was testing the LS byte of the Rune rather than
utf8 when looking for known punctionation chars.