This is a bug (well more a design limitation) in epubcheck. It validates against XHTML schemas. Nowadays HTML doesnt use XHTML schemas in any real rendering systems, making this part of epubcheck completely counterproductive. In this case the schema it validates against is complaining about the presence of value in <li> tags. This works in every real HTML rendering engine and thus calibre makes use of it. See this FAQ entry:
https://manual.calibre-ebook.com/faq...e-is-not-valid