The text has been encoded too many times, with UTF-8 character encoded as bytes, translated to HTML entities, and then XML entities (unless Perl Monks added the &).

It might be possible to reverse the mangling. The hard part is figuring out what transformations to apply, because presumbly the rest of the file does not have the problem.

Personally, I say follow the standards and decode the file based on the encoding in the xml header, and the XML entities. The corruption is the providers fault. If everyone follows the rules, then corruption becomes less likely.