As long as you know that Harmans effort was validated and therefore not theoretical. At least the 2017 version(s) were. That elevates it's status in the context of defining an objectively preferred frequency response. At the same time it was limited effort compared to their OE target, with a fixed treble band for and only 10 trained participants.
Does that make DF curves worth considering, like the one you are refencing here? If you compare multiple IE target efforts to "objectively neutral IE" you will see the modified DF curve you created is the outlier, no less between 300-3000Hz where there is a high level of agreement. The biggest difference around 3Khz where human hearing is the most sensitive per equal loudness contours.
View attachment 286559
View attachment 286560
The Harman 2017 target was validated to sound better than 30 models different models of in-ear headphones that all measured relatively poorly in "A Statistical Model that Predicts Listeners' Preference Ratings of In-Ear Headphones," but it was not validated to sound better than the 2016 IE target in a published paper. I too would have almost certainly preferred the Harman 2017 target over the other available options in the study. But that is somewhat besides the point, since my observation is not about whether the listeners, on average, prefer the 4-9 kHz energy in double blind tests, but that I do not think it is perceived as natural sounding compared to neutral speakers.
Where did you get 10 listeners from with regards to the treble in the 2017 IE target? This number was specified for the creation of the 2016 IE target, but Harman did not publish a paper on the treble adjustments for the 2017 IE target. It probably would not make sense for the same listeners to increase treble and decrease bass in the creation of a new target. One theory I have is that the reason the final treble adjustments were never published was because the new target was based on the 5128 diffuse field HRTF (5128 was released in 2017).
I never mentioned "Objectively Neutral IE Target" here. That was an old target of mine I no longer use, but some people still like it, so Kurin has it on their database. You compared that target to a "hybrid" target that is not specified in any academic literature. My current target is not too different in 3 kHz energy versus the Harman 2019 IE target, but it has more bass and less treble.
The reason I use diffuse field as a baseline, and presumably why the Harman Flat In-Room HRTF (it is not clear how this was derived) is similar to Diffuse Field with the same level of smoothing, is because timbre has been found to be associated with direction. For example, equalizing a headphone to a HATS magnitude HRTF of a speaker in an anechoic chamber at a 0 degree angle of incidence (free field target) does not result in a perceived sound source at a 0 degree angle of incidence. Since sound localization with headphones is independent of direction and considered mostly minimum phase at the eardrum, an independent of direction and minimum phase baseline HRTF seems appropriate. For the free field target, dependent on direction HRTF + independent of direction magnitude response = timbrally incorrect. This is why I consider the "Moondrop target" you cited to not be useful as it is supposed to be the eardrum response of speakers in an "ideal" listening room, which would contain directional information.
Regarding equal loudness contours: loudness was controlled when creating the Harman target. In the 2016 IE paper, preference in LF response was compared both with and without loudness normalization in tests 1 and 2, and tests 3 and 4 were repeats of tests 1 and 2 with leakage effects controlled. "The effect of loudness normalization on preferred level settings was relatively small and only statistically significant in Tests One and Two. In Tests Three and Four, the effect was not evident." It was said that the loudness model used is from "International Telecommunications Union, “Recommendation ITU-R BS.1770-3: Algorithms to measure audio programme loudness and true-peak audio level (August 2012)." "In all four tests, the playback level was adjusted to an equivalent diffuse field level of 82 dBC." Nevertheless, I do not see the relevance of equal loudness contours with regards to 3 kHz sensitivity. The level of a flat speaker is not reduced at 3 kHz (or adjusted at other frequencies even) to match 1 kHz volume, as this does not benefit neutrality. From Sean Olive: "The Harman Target has nothing to do with loudness contours as all the listening tests were done at average sound pressure levels well above where large growths in loudness in bass and treble occur."