The section about "imagined scenes" reminds me very much of the post Generalizing from One Example. I think that post and this one can be taken together to imply that, since we can be wrong about our own subjective experiences, we can be even more wrong when we try to model others' subjective experiences naively (based on our own unexamined misinterpretations). That makes the reflective process, as you describe in the conclusion, even more important.

Also, I do dream in color (as far as I know).