Now I looked, you also need to read https://metacpan.org/module/HTML::TreeBuilder#parse_file because treebuilder is interpreting those UTF-8-encoded-bytes as latin-1

I had read the latin-1 issue as having to do with file opening. Since I am passing data to parse() as a string, I thought that wouldnt apply. Thinking about it, that may be a poor assumption, but I do note that the parse() call does not mention charsets at all.

This works

Hmmm.... Ill have to spend some time looking at the changes you made.