FWIW, I would wager that any EVH "stems" you are hearing on YouTube in 2024 are not actual isolated tracks from the 1977 analog tapes. Shocker I know.
They are likely some algorithms guess at what the guitar only portion of the track is... which may or may not have some bearing on reality; I agree that real ISO tracks are often interesting and can help shed light on questions but I would not trust supposed ISO tracks of ancient classic recordings on YT. Generally think of them as more of a novelty than anything else; would not cite them as evidence of much.
Warning: Long and meandering discussion about reality of guitar tones.
It was over five years ago, so I don't know the origin. I think it was before modern stem extraction, but I don't know for sure. I agree its impossible to know their origin.
When I first got into mic recording circa 2000, I learned that even though modelers of that time were primitive (I had a Johnson J-Station), it sounded better than what I could get with an SM57 on cone, recorded into a prosumer interface of the time.
My impression is that the vast majority of the ways you can mic up cabs are wrong/inferior, compared to what can be done with mutliple mics, at the exact right position, into highest quality conversion, with best possible channel strip, etc.
In fact, what really happened
imo, is that Fractal/IRs became commonplace and that the best tones were positively selected on the internet, by clicks/likes and eventually everyone's rig started to converge to a certain produced sounds that everyone agreed was the best. This evolution mostly took place in the 2010-2020 time frame imo. A direct Fractal tone in 2024 is considerably better than what was produced in studios prior to digital.
What happened next was once this "standard" had been accepted, people went back to recreate it with real mics. There wasn't a parallel evolution of analog recording with digital direct.
So what started out was the wild west with mics on cones, with expensive desks in the 60/70/80/90s, eventually got digitized/consensualized, and then people went back and worked to recreate the best tones (consensus reached through click algorithms) that digital had reached, with real mics.
So what is real? I think this consensus that was reached with digital modeling/IRs is probably vastly different than the sounds put out the original amps and cabs, which were run through an expensive desk and tape.
IOW, what people like about modern guitar tones evolved from modeling and youtube likes, and the different devices have now converged to that, even if it is "incorrect" or vastly different than what the original gear sounded like.
So, as that relates to the TMP 5150. It may actually be closest to the real amp they modeled, or perhaps they used some creative liberty to make it "better", but the internet was having none of that, so they went back and tone matched the fractal. That is my belief.
