There is a microphone protection widget that holds the mic with four micro switches and a PCB. Should the cage around the mic touch the speaker with slightest amount of force, measurements stop. The PCB is about twice the diameter of the microphone. Even though it has an angled deflector on it, it still causes reflections back into the cone of the speaker. Because the measurements are performed in near field, this reflection causes cancellations and peaks that show up at the right frequencies.
There were two fixes. One was to remove the microphone protection which is what you see in this review thread. I am not happy to do this generally as it is risky. So I ran a second experiment where I added tons of padding around the speaker. Those results are very close to not having the protection cage. So likely that is what I will go with.
Klippel have clearly failed here. Significant reflections from their own device have unfortunately sullied and rendered all your previous tests suspect, and it is only the fact that so many measurements have been presented on different speakers, in such a short time, that a clear pattern was observed early on. Imagine some company toiling away trying to produce a perfect speaker and their Klippel rig is giving them erroneous data they can't seem to correct?
A significant re-design of the contact limit switching arrangement for the microphone and/or an increase in distance between the actual mic element and the rig mount itself should be investigated to enable quality measurements at the short distances to speakers you are employing. A extremely small proximity sensor using IR could be a better option?
When you talk about "I added tons of padding around the speaker" what do you mean? Padding to stop the mic hitting the speaker? How can that be acoustically consistent for all sizes and shapes of speakers going forward? What about side ported, side/top firing speakers? Speakers with rear tweeters etc?
Anyway, I'm glad the issue has been identified and investigated earlier rather than later. It was patently obvious there were some issues with the high frequencies in the presented speaker tests.