This has all been very enlightening, so thank you all! I was trying to compare threat models as though both systems were designed with them in mind, but it sounds like the main point I was missing was "mainstream desktop OSes were NOT designed with a threat model in mind".
For what it's worth, I'm not planning to root anytime soon. I do occasionally get fed up by the sandboxed file system (trying to move about in Termux but finding myself unable to access a decent chunk of the system), but I definitely get it more now. And I have to mirror @Xtreix 's thoughts: With stock OS, there was always a nagging sense in the back of my mind of how little control I had, but with (official, non-root) GrapheneOS, I feel much more in control and, even without sudo access, do feel like the proper owner of the device at long last. Which is why I'm willing and happy to trust the security experts here and do what they say.
Suggest you avoid derailing threads and making these kinds of attacks on our work.
For what it's worth, I don't think ekeere was criticizing the GrapheneOS security model in particular. It is definitely true that the Android security model (in the abstract) can and has been used for nefarious purposes to make it difficult to remove bloatware (I had a smartphone way back when that had an unremovable NFL app; the football app was apparently so integral to the system that I wasn't allowed to touch it). That's not your fault. That's good technology being used by bad actors to do bad things.
And I do feel like that's relevant to what I was asking about. It's easy for a newbie like me to come in here, still riding my high from switching to Linux and expecting the same of GrapheneOS. It's easy for the Android security model to get a bad rep among power users when the average person's experience with it is "No, you can't uninstall or disable Alexa, she's too important".