This is something I've not understood yet. If you rent a server somewhere to use as a private VPN endpoint, your clear IP will be pretty much the only one connecting to the server. Correlating your traffic and your clear IP to your masked IP is easy for sufficiently motivated, able actors.
Meanwhile, the main benefit of a shared VPN such as Mullvad is that many users simultaneously use the same endpoint, making it much harder to identify the user (taking only IP and traffic into account), provided they don't log your traffic.
So while having control over your endpoint is nice, how does that actually contribute anything meaningful to your privacy?
Mullvad certifiably doesn't log. Their VPN infrastructure even transitioned to RAM-only a few months back. They've been raided by the police and nothing was confiscated because there was nothing to confiscate. Obviously they have a list of registered accounts and payments, but without any connection to - well, connections.
I get what you mean though and mostly agree: There are only a few providers I trust enough to shift said trust from the ISP to them.
As mentioned in the comment you replied to: Yes, trusting a third party is a compromise. But you are also trusting a third party when renting a server for a private VPN endpoint, as well. A third party provider with probably a lot more logging going on than a trusted service such as Mullvad. While being way more exposed.
Since TOR isn't feasible for most users 24/7, trusted commercial VPNs are the next best thing when the alternative is your ISP logging everything you do.