Atmos just uses whatever speaker positions are defined in the processor, and uses whatever speakers it needs to position an object into a specific point in 3D space.
That is exactly what I said, yes. Downmixing is a generic term, regardless of the mechanism.
This is incorrect. If I have a 7.1 soundtrack being played through a 5.1 system, the rear information is discarded. We are talking about discrete information here, not matrixed. Processors don't fold discrete information from one channel to another because there could be phase issues that crop up. If you tried to play a 5.1 discrete track through a system with no center speaker, there would be no dialog. If you tell the processor there is no center speaker, THEN it splits the dialog between the two front channels.
Why would you play a 5.1 track(or any track) without telling the processor what your speaker setup is? You say information is discarded, and then contradict yourself by saying the processor downmixes so the audio isn't lost. Which is correct, yes. No audio is lost from a multi-channel soundtrack if played with fewer channels because the processor downmixes.
The way that works is different in discrete channel audio and Atmos, yes, but it happens either way. The OP's question was about real results on a real system, not about semantics.
Thanks! The Monolith HTP-1 manual
(p. 22) outlines a few ways in which channel information gets "shifted" depending on the speaker layout, but sometimes it seems the HTP-1 is making those decisions and other times it's the codec. That prompted me to look a little deeper and I was surprised there doesn't seem to be any readily-available info out there from the respective codec authors.
In my situation, I don't have room to setup surround back the proper way so I was just trying to determine if it makes more sense to put 1 speaker, 2 very small speakers (KH80), or a small soundbar right behind the sitting position. Or maybe just skip it altogether and setup my rear heights on the surround back channels. So far I've opted for a Kef 7003 because it's unobtrusive but , like many, always wondering how I can make it "better".
Very little information is available because those algorithms are proprietary(at least for Dolby). So you pretty much have to test. And yes, I agree, it is frustrating when trying to decide on your speaker layout.
I wouldn't use a single rear speaker. It was discarded from early surround formats(6.1) because sound directly behind you from a single source can sound like it's coming from in front of you. That said, if you try it, and you don't experience that effect, maybe it's fine. But that's why 2 rear speakers were standardized. A soundbar might help with this due to spreading out the source.
P.S. There are test tone files you can use to confirm behaviour, such as
7.1 TrueHD and
DTS 7.1.4 Object Emulator. Atmos 7.1.4 and 9.1.6 test files exist, but playback of Atmos in .mkv files only works on certain devices.