to quote myself about imaging:
The way I like to think about imaging starts with thinking about recording the original performance. If recorded in stereo with 2 microphones, there are two small points at which the sound is recorded.
For pinpoint imaging, you would have two correspondingly small point sources that exactly replicate the sound hitting the microphones in an anechoic room. The signal recorded by the two small points would be transmitted without any electrical or mechanical losses through equally small transducers. There would be no resonances, refraction, reflection etc. All images would appear in a narrow line between the two point sources. Basically, in the absolute ideal of the concept, this would be like listening to the performance in a dead room through two small holes in a wall...
However, listening through two small holes in a wall does not accurately reproduce the original performance (even if it accurately portrays the recorded signal). To make the sound more like a performance in real space with no wall instead of through two small holes, you have to do something to try to scale up the perceived sound. A line source would preserve much of the horizontal placement, but gain a sense of vertical scale (perhaps even scaling to larger than life); reflected sound would create a sense of the performance happening in real space; resonances can add a sense of "fullness"; etc. One of the hard parts of making speakers and why there are so many different opinions comes from trying to come to a balance of the point-source accuracy and trying to scale the imaging up to sound like the real performance.
Now, a typical studio recording makes the imaging just as much about the tastes of the people in the mastering studio as anything else.
I think Magnepan's have perhaps the closest sound to live musical instruments playing in front of me of any speakers- even if they are not the most accurate in terms of measurements.
If you are trying to recreate an PA amplified concert that was recorded through the performer's microphones, rather than microphones in the audience- I guess the most accurate set up to recreate the concert would be something similar to what they used at the show (with all the distortion of the original system).
For the most "neutral and correct" reproduction of an original studio recording, I guess you would just duplicate the mastering studio, where what it was supposed to sound like was determined (usually nearfield listening to $$$ studio monitors in a dead room)
In terms of measurements of accuracy- I think fast/smooth/controlled cumulative spectral decay and low THD is just as important as flat frequency response. Creating the same level of spl/watt across a frequency range isn't accurate if the sound coming out is not the same as the reference signal or continues to resonate after the signal ends. However, those tests can't be attempted with a cheap rat shack slp meter.