I do not have an Atmos system, but I understand about 'objects' from a computer programming perspective. So this is what I think should/ does happen. Please let me know if I've misunderstood or you have something to add.
'Objects' means that the mixer can place sounds in a 'virtual' position - i.e. an ideal location that does not have to exactly match the location of a speaker. This is the crucial difference between an old 5.1 or quad mix and an Atmos mix. In the old mixes you assign a sound channel to a particular speaker. The Atmos decoder on playback decides the best speaker(s) to play this sound from. This is similar to playing a 5.1 mix on an AVR set to 4 quad speakers - the AVR knows to create a 'phantom centre' which is actually coming from both front speakers. It's just that on Atmos the decision is not hard-coded, but is calculated depending on your speaker layout.
So... in theory the Atmos mix should closely approximate the original mix. I would therefore have expected that if you had no side speakers, the sound would come from both front and rears, to create a 'phantom side'. If 'side' sounds are always played from the fronts, this sounds like a hard-coded translation from the full Atmos mix, which does not fit in with my understanding of how Atmos should work. I wonder if the problem might be that some mix engineers do not understand Atmos. My understanding is that although you can mix to objects, you can also mix to 'beds', which seem to be much more like the old-fashioned channel assignments. Maybe the beds are only ever downmixed in a fixed fashion? I do not know if this is true, but it feels like a possibility. (I have heard/ read at least one mix engineer say they usually use beds not objects.)
'Objects' means that the mixer can place sounds in a 'virtual' position - i.e. an ideal location that does not have to exactly match the location of a speaker. This is the crucial difference between an old 5.1 or quad mix and an Atmos mix. In the old mixes you assign a sound channel to a particular speaker. The Atmos decoder on playback decides the best speaker(s) to play this sound from. This is similar to playing a 5.1 mix on an AVR set to 4 quad speakers - the AVR knows to create a 'phantom centre' which is actually coming from both front speakers. It's just that on Atmos the decision is not hard-coded, but is calculated depending on your speaker layout.
So... in theory the Atmos mix should closely approximate the original mix. I would therefore have expected that if you had no side speakers, the sound would come from both front and rears, to create a 'phantom side'. If 'side' sounds are always played from the fronts, this sounds like a hard-coded translation from the full Atmos mix, which does not fit in with my understanding of how Atmos should work. I wonder if the problem might be that some mix engineers do not understand Atmos. My understanding is that although you can mix to objects, you can also mix to 'beds', which seem to be much more like the old-fashioned channel assignments. Maybe the beds are only ever downmixed in a fixed fashion? I do not know if this is true, but it feels like a possibility. (I have heard/ read at least one mix engineer say they usually use beds not objects.)