To be blunt, you're looking at it the wrong way. You're basing all your assumptions on the "Speaker Review & Measurements" thread, and pretending it is 100% accurate. For example, the Zaph Audio ZA5.2 DIY Kit Speaker shows up on that thread with a preference rating of 5.6, but NOT recommended. Do you scratch your head and ask @amirm why that is or do you click on the link and read the review which says "There isn't lower bass. There isn't midbass. As predicted from measurements, efficiency is quite poor causing me to turn up my amplifier to very high amplification levels (1000 watts on tap) to drive them. Before you got any bass the highs were killing you and the midwoofer would start to bottom out. "You are looking at it in the wrong way.
A speaker that manage to score below 0 obviously will not be recommended. How did a speaker manage to score below 0 ?
The last 15 scored below 4.0, I would expect them not to be recommended except for the Jbl with 3.8 that this is quite a surprise.
Look at the speakers that scored above 4.0 and for the third time the list is not complete.
There are numbers of 5.x speakers that are not recommended 5-6 speakers this is quite a lot, again not all of them on the list.
The next not recommended is the Harbeth 30, receiving a 5.3 preference rating, in this review @amirm says "...we see the dip starting around 3 kHz. Getting a 3-D plot of that region gives us this complex sound intensity: I don't think this is what you want in a speaker. Put another way, it is the classic engineering work of calling a defect a feature. We can see the havoc it plays on our early window (and sound power) directivity: "
I'm not going to go down the line on each speaker not recommended with a 4+ rating. Maybe the data behind the preference rating is wrong, maybe the formula behind the preference rating is wrong, maybe the preference rating can't take into account certain speaker characteristics. Who knows. Read the review.