The more I look at this, the more I actually see a really great measuring speaker, except for that one (likely) port issue, which really drags it down. The JBL 705p actually had a very similar response, yet it received a favorable review, likely because its port null was much higher Q, and the output was better.
When I saw the first frequency response graph I was thinking the same thing. Of course, later when I read Amir's conclusion and saw this:
"As luck would have it, I started listening with one of my headphone test tracks which is bass heavy. The response here was the worst I have heard from any speaker. Bass was muted, muffled and overall fidelity just wrong. Fortunately as the playlist progressed and I listened to typical "audiophile" tracks with light string instruments and vocals, the performance was better. But have any bass come into the picture and fidelity sinks and sinks low."
I began to wonder what was going on.
Looking at the Salk web page specs, this is a tiny speaker. Yeah, Salk says you can put it on stands and use them in the intermediate field, but the Salks remind me more of nicely built desktop monitors or nearfield studio monitors, like you'd put on the far edge of a mixing board. It would have never occurred to me to expect 90db at one meter continuously from these speakers, no less 96db. I never listen that loudly in the nearfield; more like 75db average and louder peaks for movies and some classical music. This was one of the reasons why I didn't expect a great review for the AudioEngines 5+, but got a passable one anyway, because they don't strike me as going loud all that well either.
I understand, Amir, that you want to make all of the measurements you do comparable to one another. But somehow using the same criteria for measuring a tiny speaker like this Salk using similar criteria as you would a 3-way design with 8" woofers might not be the most useful approach for consumers of your reviews. Perhaps you should consider classifications based on the most likely use model. To be more specific, perhaps consider a nearfield classification, and adjust the measured loudness levels accordingly. Just a thought.
All of this thinking out loud aside, I still keep coming back to the quote from Amir above. This speaker didn't sound very good in a simple listening test.