No, I think it's a good idea. But the sample input was well balanced so I used it and included the other in (as you point out, the likely) case it wouldn't fly.

The sample also fails to specially account for the first <div/> which contains the "column" names. We could probably do something automatic with that instead of hard coding the column parsing. The OP might have thought of that already. I didn't until looking it over again now. If I have a few minutes I'll revisit it with that approach.