Maybe it's possible to bring the huge existing database into context with the new measurements by building a correction curve for the old measurements.
It would be necessary to do measurements on a number of specimen that already had run through Tylls rig and determine the difference to the new setup, separate for around ear, on ear and in ear.
I bet that also not perfectly precise, useful results could be obtained.
For the distortion measurement a microphone mounted in a flat plate could be used instead of the artificial ear.
Measurement microphones with responses up 100 kHz are obtainable, with compromises in their noise level.