As far as the .1 channel and from what I'm reading, it looks like it is missing, but where is it supposed to be? Is 2.1 really supposed to be mapped to 5.1 container or something else?
The default 2.1 has been in a 5.1 container for many years (maybe always?), as one of my systems started relying on that implementation detail since MC 21 or so (I wanted a 2.1 mixing target but three additional channels).
As I understand the unexpected/bugged behavior in this specific case, when adding one additional channel to "default" 2.1 it instead reduces the number of channels from 6 to 4 and dumps the sub channel because MC is treating 2.1 + 1 as though it were 4.0 instead of 3.1 (i.e. it treats the four channel layout as L, R, SL, SR instead of L, R, C, Sub). That is unexpected behavior in two dimensions as the choice of 2.1 suggests the user really does want a sub channel in the mix, and one wouldn't expect adding a channel to reduce the total channel count.
Now that we can specify arbitrary additional channels, it seems like the behavior of least surprise would be "make default 2.1 an actually 3 channel output, and ensure that adding channels never discards the sub channel." I'm not sure any three-channel DACs exist, so it might also just work to make 2.1 default to a four channel container in a 3.1 layout, instead of to 5.1 container, which would also solve OP's problem. Alternatively, "default 2.1 remains 2.1 in a 5.1 container and extra channels just increase channel count upwards from 6 channels" would also be relatively unsurprising, but would prevent people with 4 channel DACs (like OP) from using 2.1.
FWIW, I think most of the surprises in output format come from the fact that MC sometimes starts moving channels around as you increase channel counts, even when you're just adding extra channels to an existing mixing target. The "add a channel to 2.1 and get a 4.0 layout" outlined in this thread is one example, but as noted above if you add two channels to a 5.1 target, you get a 7.1 layout which moves SL and SR to different channels, etc. It seems like it might be worth treating extra channels as truly extra and trying to keep the locations of the "core" mixing target channels invariant.
I'll be sure to chime in when whoareyou gets a thread going!