I believe there are several missing parts here. One would be time. A measurement is a snapshot. You run a test tone through a circuit and measure what comes out the other end. To measure something like ‘space between instruments in a sound stage’, the relationships between snapshots in time is required.
The room boundaries - discussed often here - are important but not factored in when tying a meter to the end of the wire. Since no two rooms are identical, a change which may help in one room would be the detriment to another.
Another part - the human ear. This is the last stage in sound - a device at the end of every signal path that is never measured. If the mechanism used to hear a sound is not part of the measurement, what use is the measurement? Our ears may react to sound waves in such a way, with more or lease sensitivity, more or less elasticity, that some changes in electrical characteristics of sound are either attenuated or lost altogether. This is like a suspension system in a speaker - dulling out some changes and accentuating others. The human brain is a really good shelf. It is able to subconsciously shelve noises and frequencies in our environment to still hear changes in tone while filtering out fan noise or noise from lighting ballasts.
When we test fire alarm systems we do an audibility test. The recording which is played through the speakers is loaded into a hand held device. You walk throughout a space with the hand held and it compares what is being heard to what should be being played. The purpose is to determine if the words can be made out, but I think you get the idea. I consider that type of test to be closer to a real world test and you’ll find issues that cannot be identified by measuring the signal, ambient noise, or the speaker. I think the impossibility of doing the same in our listening rooms, with wires to the back sides of our eardrums, is why we reserve our judgement for a final listening test and put the results over any other test.