lynx should have an option to specify the charset of a document.
This would allow to get the correct charset when it is specified
in an external way, e.g. in "Content-Type:" in e-mail messages.

The -assume_charset option cannot be used because such HTML files
generally contain a meta http-equiv with a charset that do not match
the actual charset (I assume that there was a charset conversion when
sending the file by e-mail, but the content itself as a sequence of
characters, in particular the meta http-equiv, wasn't modified). This
is silly, but valid.