I am listening. I am just not following your narrative because the way speakers and headphones are designed, measured and practically used are fundamentally different.
Speakers excite room sound. Headphones excite sound in a small, closed cavity.
Speakers in a room can be perceived like natural sound sources. They excite the full HRTF. You can move your head to orientate in the sound field.
Headphones on the other hand do not fall in this category. They couple directly to your head and ear and transfer the sound more or less directly to your ear drum. The orientation is missing because it is not a natural sound source, hence the in-head-localization and disability to render spatiality.
These physical circumstances lead to completely different hurdles and goals when designing and measuring the system.
As already mentioned, headphones are designed to work on human heads. They rely on the closed volume to produce pressure in full fidelity. Their tuning does also involve the human pinna which does dampen and amplify certain frequencies. Same for the presence of the ear canal.
Every design decision you make - driver size, distance and angle, pad and cavity size, shape and material etc. - will interact differently with its acoustic envirnonment. Measuring on a flat plat will show a totally different response than measuring in-situ on a human head. Everything fine, as long as we are looking at only one headphone at one exact seating position. But now imagine a different headphone. Instead of transmitting the sound in a straight line right to the ear drum this one might move the driver much more to the front and let the sound enter your ear at an angle. The sound becomes more diffuse and excites the PRTF at a different angle, excites varying amplification and cancellation patterns.
Now back to the flat plate. This one could be used to simply generate the PRTF interactions by a mathematical model. But what if the excited PRTF response does change every time you reposition the headphone on the reference system, the human head or ear and cheek simulator? What if the PRTF response is different for every headphone because it has a varying acoustic design? What if the sound hits the flat plate mic mainly at 0° and adds a pressure build up due to phase overlapping? What if the angle changes and the compensation of the mic becomes more and more random incidence?
I could continue to illustrate this game for hours.
The essence is that letting a car run on a motor power testing station is a totally different experience than driving it in urban traffic or somewhere in the woods. You can try simulating specific features for those purposes. But you will never get the full and comprehensive story unless you bring the thing out to the road.
One thing we have to accept is that headphone coupling and ear simulation are based on a dynamic phenomenon which changes its (measured) impedance all the time. There is no science of absolute accuracy, just human nature and the misbelief to be able to control something that has no fixed shape. We chase after something we simply cannot control, and imagine we have mastered to get an idea of "objectivity".
Just my 2 cents.
Regards
Dreyfus