For example facebook runs an onion server for their service, they don't need to hide the service itself. So they configure their Tor relays with no anonymity on the service side (HiddenServiceSingleHopMode 1) and get better performance.
Such non-anonymous onion services can have many goals, for example: * Reducing load on Tor exit nodes * Providing users a secure, authenticated connection without depending on the CA system (assuming you got the URL through a secure channel the first time, you know only the key holder can provide service on that host). * NAT traversal for services that otherwise have no need for anonymity
30% is wrong, but there are other ways these metrics [0] are extrapolated [1]. You can see a dip in that first chart, didn't check dates to see if related.
0 - https://metrics.torproject.org/hidserv-dir-onions-seen.html 1 - https://blog.torproject.org/some-statistics-about-onions
Here is your Data about .onion websites: https://www.reddit.com/r/onions/comments/9yfwfb/my_personal_...
About 7000 .onion sites worked before, about 3100 after...
At best they took down 30% of hidden web services with published addresses at aggregator sites like Hidden Wiki.
It's not that you don't know where its hosted, it's that you don't know who is using it or where they are. That includes publishers with hosted content, even from the host itself. 5Eyes couldn't just drop a tap in front of Danial's Host and see anything useful, just intermediate nodes with no idea what was on the other end. (barring a >51% attack where they own the first and last nodes in the circuit.)