The 400Hz dip is a lot harder to explain and frankly no one has a convincing answer as of yet. Crin seems to think it's a feature of the canal itself (similar to the GRAS pinnae's sharp 9kHz dip) but I've seen no evidence to explain that yet. Especially since I've seen this dip move downward in frequency with deeper insertion IEMs like the Etymotics.
We talked about it already, but I'd really make a distinction between the overall alteration of the SPL ratio below 800Hz or so and above,
which is a product of the 5128 being more representative of the average human ear canal (dotted white trace), and the 400Hz dip specifically (red circle) - which is particularly prevalent with Crin's measurements.
Just like you I'd like to know more about its origin, but intuitively the first thing I'd do, given how the 5128 is constructed (is it me or the coupler is only attached to the pinna / canal silicone combo, and then only the latter attached to the HATS - and in the case of Crinacle to nothing ?), would be to try to securely couple the coupler / pinna combo to a stable mount, or use a massive load of putty to couple it all together.
What we need is in-canal measurements of 711 couplers, 4620 couplers, and human ears to compare and see if this dip shows up at all on humans.
We do have some

(CSGlinux). cf above.
This would beg the question if BK5128 measurements are misleading
The rocking modes around 100-200Hz probably.
The 400Hz dip possibly. I have seen neither in the few in situ IEM measurements I've looked at, but it's possible still that in some specific situations it happens in humans as well.
The overall alteration between below 800Hz or so and above, nope, that's just a better representation of the average ear canal.
At really high frequencies individual variation starts to get very high (cf link above) and you have to combine it with insertion depth (and different IEMs react differently to a similar variation in insertion depth), so it's going to be a bit of a crapshoot anyway.
Since (cf article linked above) the difference between two individuals in terms of acoustic impedance in the 20-5000kHz range seems to me to possibly be just as important as the difference between 711 and 5128 couplers, I'd also wonder if we could see quite important differences between individuals in that range as well.