For the first two, PDF metadata extraction is mainly for academic papers. Any document can be printed as a PDF, but that doesn't mean that there's metadata in any sort of standard form that could be extracted. So in cases like those, you'll want to right-click on the document, choose Create Parent, and then fill in metadata manually.

The third isn't a PDF, but it's the same problem — it's just a random document that was exported to HTML. Zotero will detect the DOIs in the bibliography and offer to import metadata for those, but it can't get metadata for the page itself. Zotero has support for most mainstream journals, but for a journal like this where it's really just exported webpages, they would need to embed standard metadata into the HTML for Zotero and other tools to be able to extract it. This is a rare case where you'll actually get better automatic metadata by pasting the title into Google Scholar and saving to Zotero from there, which will get you most of the metadata.