I believe our host has stated in the past that one of the major differences between a mixing and a mastering monitor would be intended placement and listening position. Mix rooms (and recording control rooms) are often constrained in space such that the engineer will almost always be working in the near field, or at most mid field, whereas mastering is best done in the mid or far field. Additionally, mixing often requires use of more equipment—and therefore more furniture—than mastering, so having greater in-built boundary compensation features may be desirable.
As for everything else, I would posit that flat anechoic magnitude response, phase coherence, low distortion, minimized resonances, constant directivity, good headroom, and high peak SPL should be desired for all monitors—and, by extension, all indoor loudspeaker systems. Dispersion may be down to individual engineer preference, but I would imagine no one wants a laser beam and no one wants an omnidirectional speaker.
I can understand the appeal of a “reference” monitor for mastering, but as magnitude response (i.e. tonality) is the most salient aspect of a near field monitor, one should be able to satisfactorily approximate one monitor on another using EQ—especially if the latter meets most of the criteria specified above. You can even use room impulses and convolution to hear as close an approximation to what the mix engineer heard as possible, plus emulate any number of other playback systems and locations. Only one loudspeaker monitoring system should be necessary, but I would add a headphone system, the listening experience of which can’t be emulated on loudspeakers.