-there are obviously some headphones with massive resonances or nulls that can hardly be just compensated with convolution without creating a bunch of other issues.
-then as said, if distortions are high enough, they will make a noticeable difference. not much we can do about that, but logically FR is still a bigger deal as it involves all levels of loudness while distortions are usually -30 -40dB or lower.
-for generic compensations, there are simply the manufacturing tolerances. a few dB here and there aren't that rare.
-but there is another parameter to take into account. even if they succeed in EQing 2 headphones the same way, that EQ is based on a microphone on some dummy head. just by using our head instead, we can expect some measure of variations from headphone to headphone. how we will end up placing the headphone for comfort the size of the head and clamping force resulting of it, if one headphone has a small diaphragm and the other has a big one or if one is angled... then the signal bouncing on our ears will create differences, but not necessarily the ones captured by the dummy head. it should be fairly close because, well, we're still humans in shape, but all those reasons can explain why someone may not get 2 headphones to really sound close from one another.