Some sites are so broken that the only way to parse some of their HTML is to use regexps or substring searches on the source code. Maybe this tool can handle that through extensions.