So I'd argue it makes more sense to first either get a decent alternative OS running on a proprietary piece of hardware (which is almost impossibly hard) or get an open piece of hardware running the only thing that comes close to an open-source OS first (so users will actually buy it because they can actually use it as their daily driver).
It doesn't make any sense for any project without billions of dollars backing it to attempt the 100% purist approach from the beginning. Even Mozilla wasn't able to pull it off, so how and why should anyone else succeed?
Mozilla did not try a purist approach. Their Boot2Gecko was built on AOSP downstream kernels and proprietary drivers. (Then again, there was no such thing as a truly 'purist' device back then, even in non-AOSP land. So they just did what could work at the time.) If you want to explore that kind of approach, there's things like UBPorts and Droidian. They might even make sense, if they end up being able to function as AOSP-compatible Generic System Images (GSI's) and run on mostly any Project Treble-capable device. Make no mistake though, that's a lot harder than doing everything properly in upstream and restricting one's attention to hardware that's compatible with that approach. If only because you're not forced to keep up with churn in the AOSP-native interfaces.
I went from Maemo (Debian based) to AOSP Android when my n900 no longer worked. I've never experienced such a user-hostile platform as Android-- you even have to expend effort just to gain and retain root on a device you nominally own. And, I don't trust anything in the Android ecosystem to not be spying and tracking.