I guess it depends on what you mean by "response". If you mean estimated in room response(ie PIR), I agree, as the currently generated PIR (I'm pretty sure) is incorrect for dipole speakers. If you mean the anechoic response, I disagree. You want to know how the speaker radiates in three dimensions free from all room interaction, which is what the NFS gives. From that, you can calculate room and boundary interactions at various distances with various wall/ceiling/floor reflectivity. If you measure in a non-anechoic environment, that measurement is only valid for that particular boundary distance, boundary reflectivity, and mlp distance combination.
If anything, anechoic measurements are even more important for speakers that interact with the room to a greater degree. They allow us to estimate the response in many different types of rooms.
The only exception to this is when the radiation pattern of the speaker itself is altered from its free form anechoic pattern(ie in wall/soffit mounted speakers). In that case, you want to measure it's "actual anechoic" configuration(ie on an IB). You wouldn't measure a normal monopole with the driver outside of the baffle.
You could argue that this is partly true with panel speakers(due to the floor interaction), and I think you'd be right, but luckily this speaker was measured on a solid platform(ie not suspended in air) and the error is in the vertical plane. The NFS measurements are still the gold standard for speakers like the LRS or any other panel speakers.
Where the error comes in, is in the interpretation. This is an area where my mind has changed. When first viewing these measurements, I judged them as horrible, but that's because I was judging them by what good(highly preferable) monopole measurements are supposed to look like. I was judging them by how close they came to that flat on axis, smoothing increasing directivity standard. This was (imo) wrong on my part. I don't think we really know what measurements of speakers like this are supposed to look like, so we really have no conclusive way to say whether or not these measurements are good or bad. Unfortunately, this means that measurements for speakers like this really aren't all that useful. Not useful doesn't mean wrong, though.
What we need is many more Klippel NFS measurements of speakers like this. Then we need many blind listening tests that we can use to start to understand what a SOTA panel speaker measurement should look like. This will also start to make these speakers even better, as designers will have a much clearer target to aim for. Until we have these measurements and listening tests though, these measurements don't have much value.
This is not an issue with the anechoic measurements, though. This is an issue with trying to use those measurements to predict performance.
I agree -- the anechoic measurements themselves, as opposed to some of the inferences made from them, are valid and potentially very useful. In fact, when Amir was first asked to measure the LRS, I said that I hoped the measurements would shed some light on the fact that planar dipoles typically sound a good deal better than the conventional measurements would suggest, which accords with your goal of finding better criterion by which to assess them.
There sure are some significant challenges here. How does one interpret the measurements of a speaker that (in the shorter models, anyway) is listened to partly in the near and partly in the far field, with the transition depending on listening distance?
How do you develop a quality metric for something like interaction with room modes, or the effect of polar pattern on the reverberant field and the fact that it shifts optimal RT60 to something closer to that of the typical living room?
How does the elimination of the baffle step and uniformity of the polar pattern (in the higher end models, anyway) affect subjective impressions? (Linkwitz felt that uniform dispersion was crucial to the sense of space.)
There are so many variables, as there wouldn't be if we were comparing loudspeakers within the same family, that I can often tell more about the sound of a loudspeaker from knowing whether it's a dynamic or planar magnetic or ESL or horn than I can by looking at the measurements. That I think makes it more challenging to develop metrics that apply across families, e.g., the tendency to excite room modes -- all monopoles are similar in that respect, and so are all dipoles, but they aren't similar to one another. (A related example -- both monopoles and dipoles suffer from boundary effects, but a monopole can be positioned with respect to the front and side wall to minimize the effect, while a dipole can't, but doesn't suffer from floor and ceiling bounce the way a monopole does -- in fact, a line source requires it.)
I've become close to the guys at Magnepan in recent years, to the point at which I've been collaborating on the design of their forthcoming speaker. In the course of that I've learned a lot about the measurement and listening methodologies that they use, which are the product of 50 years of experimentation and refinement, backed by a robust commitment to blind testing. There is so much lore there that I wouldn't know where to begin even if I were privy to it all.
Response measurements are an example of the issues that they've learned to overcome. It's difficult to make a gated quasi-anechoic measurement of a tall line source, and in any case, measuring in the near field would lead to an inaccurate measurement of bass response. So in practice, multiple measurements are made -- gated measurements to tune on-axis response, in-room measurements at the listening position to tune bass response, polar measurements to measure power response, etc. And then the known modal behavior of the room and the boundary reflection from the front wall has to be taken into account as well to yield a response that is less dependent on the specific room (and since it still remains somewhat dependent owing to acoustical baffle extension, measurements and listening have to be conducted in rooms of varying size as well). Needless to say, it would be a boon to be able to do that from anechoic measurements of the sound field!
So it's worthwhile but hard to bottle this, in part because they're beginning with empirical results (listening backed by blind A/B testing with non expert and expert listening panels) and working backwards to find the salient measurements and design techniques." But I think it would be wonderful if objective criteria could be established, as they have been for monopoles -- something that would probably require modifying the criteria for the assessment of monopoles as well, to account for factors such as their differing interactions with the room.