That would be a poor assumption. How can it distinguish a single user's reader from a shared service like NewsBlur, which will request a feed on behalf of hundreds of users? The server knows that some aggregator is pulling its feed, but it has no way to tie that to a user.

There are other ways, though: images in the feed that call back to the server, for example.

True, but it is quite possible that is you yourself.

If you selfhost your reader, you control the logs. If you use a reader, the server admin might now what feeds you have. But the server who hosts the feed (the blog) has no idea, tracking doesn't work easily - that is still an improvement.

Note "feeds" is plural. There's no way for a server admin to get a list of other sites you subscribe to, unlike a centralized app like Google Reader.