After auditioning various speakers, I definitely agree that specs don't always match the actual sound for some mysterious reaons.
Aesthetics, Brand Name, Prefessional and Peer Reviews also play into the buying decision.
There are 2 major differences I've found in the sound quality of the speakers.
The first and most obvious one is the bass output. This one is easy to see on the frequency response graph.
If you listen to towers that can't play flat to 30Hz in-room-response and you don't use a subwoofer, the sound may just turn you off.
Speakers that have built-in subs like DefTech BP7001/7000 will be able to dish out the 30Hz easily even if you don't add an external sub; it's almost like you get the sub whether you want it or not.
The second difference is the midrange resolution - how detail and clear the midrange is. Some speakers will be able to play instruments clearer than others. Can we see this in the frequency response graph?
Is it how accurate the speaker is on both the on-axis and off-axis FR?
In my limited experience, I postulate (but not theorize)

that a speaker with a 5-point average Freq Resp On/Off Axis from 200Hz-10kHz of < +/- 3.0 dB (on Soundstage Anechoic chamber or at least on Home Theater Mag) will have a satisfying midrange resolution.
The imaging/soundstage is kind of over-rated IMO. Perhaps only the worst speakers have issues here IMO.