I just downloaded the data and script and got not well-formed (invalid token) at line 1163, column 17, byte 41545. What I find there is a ³ - so it looks as though the problem is insufficient encoding.

I'm looking at the XML::Twig POD right now to see if I can dredge up anything fitting. Passing keep_encoding => 1 and/or input_filter => 'safe' to new() didn't help.

Makeshifts last the longest.