The accuracy of your simulation depends on how far down you want to go into the simulation process. On the fundamental level, it's actually pretty easy to simulate the physical processes of thermionic emission, circuit simulation, dynamic impedances, EMF fields, magnetic saturation, etc... and then just build up from there.
The problem is, it's not yet practical to do all that in real-time.
The good news is you can measure and parameterize many of those things and get reasonably close.
The bad news is that it's impossible to do that for every amp combination.
The other bad news (for modeling amps) is that you can still buy a Traynor for $400 and it won't take a shit on you halfway through tour.
The OED lists the following definition with examples going back almost 60 years:
"model, v. 10. trans a. To devise a (usually mathematical) model or simplified description of (a phenomenon, system, etc.)."
(The use of "simulate" to mean a similar thing dates from <70 years ago, so neither usage is hugely older than the other.)
This is not "weird"; it's a standard term in applied mathematics and other fields. There's even a monthly journal Applied Mathematical Modelling: Simulation and Computation for Engineering and Environmental Systems.
The transformer used to match the high impedance tube output to the low impedance speaker is also a factor. When that transformer reaches the magnetic saturation point, the bandwidth of the transformer will narrow, depending on the degree of magnetic saturation.
The speaker is a factor as well. It can be driven into breakup, though I've never tried it.