Be aware that measuring FR and THD is just that: measuring FR and THD: it's not a measurement of how good or bad something sounds. It's the equivalent of measuring a car's quality by looking at 0-100km times: it doesn't even tell us how fast a car will be.

There is of course room for individual preference, and one's hearing and the physical headphone fit play a role.

But essentially: yes, to a reasonable degree we absolutely can tell how good or bad a set of headphones will sound via measurements.

    It's the equivalent of measuring a car's quality 
    by looking at 0-100km times

The "quality" of a car is the sum of hundreds if not thousands of factors.

Audio reproduction has basically a single input, and the output can be well described using a small number of metrics.

We know what sounds good to people. It would be a miracle if we didn't, given the fact that the business and science of audio reproduction have been around for 125+ years. It has been studied and is something of a solved problem.

Sean Olive and others have published some great research that's worth reading. The TL;DR is that headphone listeners prefer accurate sound reproduction that is fit to what is sometimes called the Harman Curve, which in turn correlates pretty well with the Fletcher-Munson equal loudness curve.

It's interesting that these debates pop up around audio but not video. We know what "looks good" in a television: objectively accurate color and motion reproduction. This can all be easily measured. Audio can easily be measured as well, and yet....

If FR and THD measurements are bad, at least the sound is not neutral.