First, I'm sure you could do that. And somebody would like it. But do you really think a significant number of people would prefer it to the Harman curve?
I have no idea and that's part of why I find the idea interesting.
I've seen enough to give me the idea that a preference curve is a rather slippery idea.
Part of it goes back to
purposes and statistics. If you are in the business of designing headphone models for mass production then it makes sense to measure the preferences of your target market and design models for some kind of central tendency within the data. That's not a very convincing basis for me to believe that I will experience good sound by choosing a product that conforms to what that manufacturer deduced from its tests. Maybe I'm quite unlike the target market, or I'm in the target market but two sigmas away from the mean in some way, or I have defective hearing or something.
If we had the database of raw test results and can identify our own in that set then it might be quite educational.
The preference curve is a very appealing idea because if such a thing really exists that with decent probability predicts my preference then that simplifies buying equipment enormously.
Otoh, I bought the Crinacle Zero IEMs on the basis of Amir's review which showed excellent conformance and I mixed an episode of a podcast using them and the result was terrible on anything I played it back on (in the car, on a portable BT speaker, only my desktop system with Genelec monitors, only bone conduction, Koss and Sennheiser phones).
There's a difference between data and information. Data is neutral. Data simply is, like stones, rain or stars or other unintentional objects. To find information in it we have to ask questions that relate to our
purposes and then see if the data can
inform those questions. Or something like that. But I'm sure data and information are qualitatively different. And Harman's curve relates to Harman's
purposes.