You misunderstand. In the graph what's shown is the response of a
pressure microphone, which are used in specific scenarios (like closed box tests). The microphone we use for measurement are free field and do not have that kind of aberration. There's a difference in construction which accounts for the change in response. Pressure microphones are also more directionally sensitive than free field mics.
This is perfect CD, taken from Geddes' paper.
View attachment 295059
Another "more feasible" version of the same.
View attachment 295060
Here's his model on a great loudspeaker's polar pattern:
View attachment 295061
I think you're overstating the case when you talk about the "definition" of CD in terms of DI, same as you do for dipole or cardioid speakers "interacting with the room less" (dipoles excite pressure minima where monopoles excite maxima, while cardioid speakers have limited relevance in the modal region of small rooms). Same goes for this overstatement:
Unless there have been some kind of preference studies done that I'm unaware of, there are no separate target on axis curves for those speakers.
Almost none of this pertains to a "psychoacoustical perspective" about constant vs. rising DI you wrote about originally.