In his essay titled, "Horses for Courses," John Marks points out that an all-around test that is representative of everything a car (or audio system) is asked to do may be preferable to just acceleration times (or distortion measurements). I am also reminded of how computer benchmarks show one winner for some tests and may show another winner for a different test that emphasizes I/O data transfer instead of CPU speed, for example. Each of these makes the term Horses for Courses a relevant subject.
I agree with the essay that there is no obvious test for audio equipment that is both objective and all-encompassing. I think it would be at least entertaining, however, to see data taken in a recording venue (such as a concert hall) that compares a live recording with that same recording played back through a few audiophile systems, and a very low cost system or two as well. Equalizing the recorded levels as well as possible to minimize differences, but then looking at the live recording minus the playback recording for each system, would give some indication of the feasibility of measuring overall performance. Some systems may win at chamber music or solo classical guitar, while others may win at hard rock or full symphony orchestra. Horses for Courses.