It wasn't meant to handle unescaped <, >, ' or " characters. Your sample data didn't mention that as a problem. For the general case you might try referring to merlyn's column The Wrong Parser for the Right Reasons.