Have only just discovered this thread and the reported issue.
I cannot see why it matters whether the assumed speed of sound is incorrect, provided it is applied equally across all channels.
The only circumstance I can see this matters is, where one is applying some degree of post-processing, such as Audyssey, and the processing introduces differential signal delays per channel which needs to be compensated for based on speed-of-sound assumptions.
I made similar points before, though we don't really know if it is applied equally across all channels, even during the initial calibration. For example, if D+M use the channel with the longest delay as reference, and time aligned the other channels to it, then the speed error may matter.
Again, in my thinking, we all know the mic measures delay, not distance, so if say, Denon (or Marantz) accepts the delay information directly (as they should, but that's just my opinion) then there will be no error due to the incorrect speed as there wouldn't be conversion. In that case conversion will only take place when Denon converts it for the interface that displays distance.
However, if Denon only accepts distance, then it would have to convert the measured delay into distance and in that case, whether we run calibration with the app or directly on the AVR, the delay info will have to be converted to distance first and then back to delay for Audyssey to do the filters and time alignment as required. That's the lengthy discussion I had with Nutty667. He thinks it is better to do that double conversion so that there would be one "path" (what he called), whereas I think it is logically silly to convert for Audyssey to do its thing, when there is no need for it, but I agreed with him to disagree that there are just different ways to do things and we may not prefer the same ways.
In the end, Nutty667 is most likely correct that Denon only accepts distance and will then convert it to delay for Audyssey to do the time alignment and filters, whether I think it is silly, or not. If that's the case, then the speed error will matter unless as you said, it is applied equally across all channels, but that's another unknown and I wouldn't like to take a guess any more than I have had.