It took me a while, years actually, after I got into headphones before I understood what exactly I feel when it comes to slam. For me, the most direct instance of slam is the "chest punch" that people feel with speakers playing a sudden loud note. When the bass note hits on a nice pair of speakers, there's this sudden physical feeling of pressure in the sternum, probably from some chest cavity resonance being excited. Headphones obviously can't have chest punch, so the closest thing they can produce is what I call "throat punch" where it's the same sort of feeling as chest punch but localized to the upper throat and nasal passages instead of the chest. It's the same sudden physical feeling of pressure or tightness in the throat and sometimes the eye sockets when notes (usually bass notes) hit.
The physiological phenomenon I've seen that's closest to what I hear is tympanic response, the response of muscles in the ear to sudden increases in loudness. It's part of the auditory flinch response, like the flinch I feel when a trumpeter plays a note at full blast. Since tympanic response is pretty slow, it makes the biggest effect in bass notes since the period of those frequencies is longer. On mids and treble, that 'slam' effect is more like this sort of "liveliness" in the sound. It sounds weird to say, but I would say that I technically don't hear slam, I feel slam. I don't try to hear if a headphone slams, I ask myself "do I feel the hit of bass notes in my throat?" If slam is related to tympanic response, then it makes sense that different people will perceive it differently, because people have different flinch tolerances.
My theory is that slam is related to the magnitude of the crest factor produced by the headphone during transients. It's not just about "static" magnitude in the bass as seen in the FR graph, but about behavior during transients. The more a headphone overshoots the intended crest in the signal, the stronger the "slam" in that particular set of frequencies. You can recreate this using dynamic EQ; use an expanding filter and set the filter to use auto-thresholding. That will make the DEQ plugin (e.g. FabFilter ProQ 3) look for transients in the filter's frequency range and dynamically increase the boost when a transient occurs. This causes the transient to "overshoot" its intended level and effectively increases the crest factor of the transient in those frequencies (and overall). This produces slam. In fact, it's possible to make a headphone less bassy and yet more slammy at the same time by cutting the bass via parametric EQ then adding an expanding dynamic EQ filter on top.
I don't think that slam directly lies in the frequency response, though some aspects of the FR will affect slam. Since dynamic EQ is a nonlinear EQ process (unlike standard parametric or graphic EQs), I strongly suspect that the actual metric behind slam is also not linear, in the realm of distortion rather than frequency magnitude response. But what kind of distortion it is, and how to capture it, is something that I don't have enough know-how to answer. Dan Clark has made some comments on this matter and overshoot, but he has no incentive as a manufacturer to spill the beans on the whole story.