To me, a really great speaker would be designed to output the original signal as faithfully as possible. Then, theoretically anyway, if you have these speakers in a good room and set up correctly, and if the system upstream is also accurately reproducing the original signal, you will get a good facsimile of the original recording.
I believe good speaker designers have identified and analyzed shortcomings with existing designs and made efforts to correct these deficiencies. This has resulted in many designs that do perform better than other previous designs, at least in those areas the engineer was addressing. It is perhaps our sensitivity, or lack of sensitivity, to these areas, as well as those not specifically addressed by the designer, that lead to our brand or model preferences. And even this is limited to what we have actually been able to hear (has anyone heard them all?). So we end up with speakers that surpass all other designs, but usually only in one or two areas.
For instance, I currently think that B&W makes some of the most accurate [in many areas] drivers ever produced. You could definitely build a great system around some of their D series speakers. But the sound just doesn't hit my ears right. Does that make me or them wrong? And then nobody does lifelike dynamics and resolution better than Wilson. If careful, someone could have a great system around these too. But not me.
For me, the speaker system I have heard with the most potential with the fewest compromises is the Thiel CS-7.2. If set up correctly, this speaker system has resolution, neutrality, soundstage and imaging to rival the best, but without the shortcomings (comparitively) I am sensitive to of other designs I have heard. Definitely one of the best, if not the best speakers available.