The first part is correct: imports are stored in the same way as regular entities, but what is different is the $entity_header vs $entity_import_header.
The header tables are what allows us to say “Here’s an entity of this specific type, and here’s where you can find the latest revision (current state)”.
By having separate headers, we can identify and mark imports as such on the website. Once validated, we move the $entity_import_header entry into $entity_header and voilà! consider it imported.
The link_import table looks to me like a way to say “this item was imported with website ABC where it had ID #1234 and has been fully imported into this BB entity”.
For example, if we import the Bookogs DB tomorrow (I wish!) and store all that information, we can later re-import the database with images, and know which item has been imported (or not) into which entity.
Sorry, I forgot to mention this feature is currently being implemented, and not currently on the website.
For more on this topic, see Author aliases - no method of crediting works under pseudonym/s