The file self-describes (in the first line) as being HTML, so not using characters that don't work in HTML seems like a sensible requirement. Strange that nobody noticed and fixed it, earlier.
Perhaps browsers of yore (heh, I was there so that sounds strange) did render the elements in question since they were less strict?
HTML1.1, HTML2, etc weren't less strict, if anything they were more. They only consumed tags that could be rendered, since there wasn't a concept of a DOM tree or anything like that. They just worked forward rendering as they went. CSS and JavaScript introduced the need for a DOM tree and mutable state, in which case HTML renderers treated HTML docs more like XML descriptor files than SGML.
Perhaps browsers of yore (heh, I was there so that sounds strange) did render the elements in question since they were less strict?