For speakers the 45 degree angle and near field in a treated studio would come close to the plot you posted.
Should you use a HATS to measure the sound at the listening position then neither DF nor FF correction are correct nor is the plot correct as the HATS response will differ from the overly smoothed drawing.
So the compensation needed to calculate what the sound was approx 10cm before it entered your ear (which is what you hear) will differ IF you want to accurately know what that was.
However, in headphones the sound does not come from both sides, it does not come in from a 45 degree angle, it does not interact with the room but comes from a very short distance in a small (somewhat or completely) sealed 'chamber' so different compensation is needed to deduce what sound would come from the driver and what you would hear.
Trained listeners are needed here. A reference speaker setup would be needed here that is confirmed to reproduce faithfully in a treated room (studio) with speakers at a certain distance and music played at a reference level (say 80 to 85dB average).
You measure with a specific HATS and create a compensation curve so that what comes out of the HATS is similar in FR to what a linear mic would pick up.
Now you have a target curve... for THAT specific HATS in that particular room with particular speakers at a specific angle with specific room conditions (treatment, speaker placement, console, chairs, windows etc.)
That will differ from the plot you posted and won't be 'smooth'.
Difficult part... it won't represent every other studio, not even when the same speakers are used.
So more research ... more measurements, more listening, more plots with the same HATS.
More averaging of the compensation etc.
Now there you will see something looking like the plot you posted. It won't fit every studio/circumstance but is better than nothing.
That last bit is where a single compensation curve comes in... it is somewhat accurate in general usage.
Now... headphones. Angle and circumstances differ.
Trained listeners, could even use a really good studio and do the comparison there.
Same (measured) average SPL, listen, compare, EQ the headphone (most far more wonky than speakers) and get the sound to be the same.
The measured response is how the trained listener hears the 2 situations. The HATS response is known. We can calculate the needed correction.
That curve will differ from the one made with the speakers.
problem... put another headphone on the head and repeat... somewhat different compensation is needed. Repeat with 10 headphones.. all result in different compensation curves. move the headphone a cm and different yet again.
Now this is impossible to use so we need to look at the average and use that to get an idea of how it 'sounds' to a trained listener compared to a reference with speakers in a well treated (one may hope) room. That compensation curve will be smooth and 'on average' correct up to a few kHz.
You can do a similar 'thing' with the same excellent speakers in a 'good' emulation of a somewhat treated domestic room.
Use a HATS and trained and untrained listeners (the public you want to sell to) and obtain a somewhat different average correction.
Bass and treble is often an issue so let listeners play with tone controls that are fixed in frequency response but can control dB's.
Again, heavily smooth the average measured response on a specific HATS and you obtain a somewhat correct response up to a few kHz.
One can say 8kHz tops or 10kHz tops but in reality it will be lower than that.
One can differentiate the responses for certain groups of listeners. This is nice as research.
Guess what... Harman research...
Now we need to sell headphones to the general public so average the set amount of bass and treble from the research and come up with something that is preferred on average (is not the same as professional in studio circumstances) and you end up with the Harman curve for 1 specific HATS.
Using that specific HATS you can get a good grasp of what the tonal balance will be from bass to say 6kHz or so (one can always claim 8kHz or even 10kHz) provided a good seal is obtained. The latter is not always the same and the positioning is not always the same either.
So headphone measurements... it is a somewhat averaged measured response that can give a clue but is definitely not an exact response.
This is why, even with accurate EQ based on such measurements, will not result in the exact same sound but on average, in perfect conditions, will make headphones sound very similar in tonal balance but can still differ above say 6kHz where the most gremlins will be.
conclusion: headphone measurements is not an exact science. FR measured by a HATS is an indication at most for tonal balance and EQ should not be based on an 'exact' opposite of the measured response on one specific HATS in one specific circumstance using one specific 'correction including tonal preference IF you want something 'exact'.
That said you can make headphones sound more similar to a specific target.
And guess what... on average people seem to prefer a scientifically obtained target. Others may prefer a different 'target'.
Harman response is just that... for the average public (that likes some bass boost for whatever reason) when those headphones are measured on that specif HATS, or in the Harman case a specific GRAS flatbed fixture with a specific pinnna and coupler. When every one would use it.
But there are many different test fixtures with different responses..... the horror.
Take measurements of headphones for what they are... not as an absolute truth. Regardless who does the measurements with what fixture.
You want to be close to Harman ? use the same fixture.
