I'm not sure why pdf2html converted those to soft hyphens. I guess that's how the PDF was made?
Manually fixing them worked, but of course it was a pain (turn on debugging, convert PDF to epub, grab the html output, modify that, import it, convert html to epub, clean up poor PDF line unwrapping in the epub with Sigil). I know it's my fault for wanting to convert PDF, but I hate hate hate PDF as a format