There are PDFs where the text is behind an image - you can select it but as far as Calibre's PDF conversion is concerned it cannot see it as anything but an image. If you find a book where you can convert it to PDF and select the ISBN as text in Calibre's ebook viewer then it might be a case the plugin should be catching. As I said above, click on the View log button to see what numbers it found if any, it might be rejecting them as invalid ISBNs or it may not be finding the numbers at all. It would be fairly unlikely that you have found another valid ISBN combination that is being rejected by the plugin due to something I can control (such as variations of emdash) but anything is possible.
As for djvu's, I don't own any so perhaps someone else can comment, but I think I read somewhere that Calibre didn't use to convert them anyways. If Calibre cannot convert a format to epub, then this plugin will not work with it.