A speaker responds to the voltage on its terminals. If you model everything accurately and then simply amplify that signal and send it to a speaker the results will be the same. The hard parts are modeling it accurately and amplifying it correctly.
Modeling it accurately means modeling the speaker impedance which is often not known and accurately modeling the I-V relationship of the power tubes. The latter is extremely difficult and requires a lot of processing power. Inexpensive products use waveshaping and EQ approaches. We use nonlinear ODEs and iterative solvers.
Amplifying it correctly means an amplifier with significant power reserves. Most of these small, cheap Class-D amps simply don't have the power reserves to replicate a cranked 100W tube amp. The transient response is lost because the amplifier runs out of energy. They aren't designed for these sorts of applications. They're meant for low-cost consumer applications.
High-end solid-state amps, whether Class-AB, Class-D, Class-G, etc. (i.e. Crown, QSC, etc.) have the requisite energy reserves and I bet anyone would be hard-pressed to tell the difference in an A/B test (assuming the speaker impedance were set correctly).
I've done tests comparing various 100W amps using a Crown K2 and, a Matrix (something, forget the actual model but it was 1000W+) . The differences were negligible IMO. In fact, I could tweak the speaker impedance curve and end up with something that actually sounded better.
It really depends on your application. At loud stage volumes an inexpensive Class-D power amp isn't the right tool for the job. In a small club application then it's probably fine. Don't confuse misapplication with some nebulous physical shortcoming of the various technologies.