It pulls URLs from your history, and downloads them later. It doesn't download what you see, so if you're logged in somewhere, it won't be able to save that; or if content changes between when you see it and it gets scraped, it won't know.
The use case you raise makes a lot of sense, especially on sites with paywalls or where you have paid accounts.
I haven't explored this route personally yet, but it may be possible to copy cookies from your browser cookie jar over if that would satisfy your use case.
