I hate to be a stickler but I'm afraid to say that metric you proposed, about the difference between the early reflections curve and the listening window, would likely not be unambiguously preferred...
For example, some research suggests that a speaker with uneven dispersion (and thus, an uneven early reflections curve) will be preferred to a speaker with better off-axis performance if the former has
wider dispersion. I've posted this previously on this forum but one of Toole's studies put up three speakers with even (for the time) on axis performance and varying off axis performance.
View attachment 45309
In this example, the Quad ESL clearly has the most even off axis response that is most timbrally similar to the on axis curve. The KEF and Rega both show significant differences. We'd probably call that off-axis performance straight up bad if these were modern speakers.
And yet, the Quad ends up worst in listening tests.
View attachment 45311
We saw something similar with the blind test comparison of the JBL M2 vs Revel Salon2. The Salon won by a considerable margin despite the M2's on axis and off axis performance both being much 'prettier' than the Salons.
Moreover, other research tells us that mixing engineers have a greater preference for narrow directivity designs and/or fewer sidewall reflections while doing their work than do recreational listeners and even mastering engineers, so that's another confounding factor. And individual preferences still exist in this regard.
Also note a timbrally balanced speaker with tall, narrow peaks and dips is likely to perform better than a speaker with a big shallow scoops across various frequencies.
While I'd love to see a universal preference metric as much as everyone, creating one really is a monumental task. Not saying it can't be done though or that we can't get to a useful approximation, just explaining some of the hurdles. With DACs and Amps we have some very clear goals for performance, and we largely know the limits of what's audible. With speakers, we are chasing a moving target.
Id still like to see information like standard deviation from a target curve and such, but we need to have an understanding that diminishing returns and preference have a role to play.
I'm sure Amir's conversation with Sean Olive will be fruitful though.