When we wrote the current version of our product back in 1995, someone else wrote our parser. Several years ago, I went through the parser from stem to stern, cleaning it up so that it became much better behaved. Unfortunately, there were a few seldom-used functions in the parser that I pretty much papered over rather than converting them to use the new routines, because it wasn't worth the time.
I'm getting to figure out how to convert those now, because it's easier to convert them than it is to figure out how to paper over them again. *sigh*
On the other hand, when this is done, things should be very sane in the parser. I hope.