We tried to lookup the font metrics using the font name as-is, which didn't work since the PDF file in question has non-embedded fonts with names that include commas.
Hence the font names need to be normalized here as well, similar to elsewhere in the font code.