MSOCS,
You are absolutely right, but we are talking about different things. No doubt about great value of LVC, and I am its consistent and long-term advocate (and at this "campus" since this term was born). Moreover, what you called correlation is actually a monotonic function (thus, the correlation is indeed 100%). The more you train in such a way, the better you would behave in a real situation, no matter what would be the win/loss ratio.
I am only challenging numbers because the entire adversary system contains thousands of parameters and hundreds of links, and in the West just a few dozens are known (of old MiG, Su, S-200/300 and some other already obsolete systems that Eastern European countries were happy to deliver to new "landlords"). "Validation" in local conflicts against some stand-alone systems and poorly trained military is quite questionable.