Violent destruction is not dis-assembly. The event was chaotic. How many models of chaotic events are modeled with any reliability?
All you need to determine is a rough range of models in which rapid progressive collapse occurs. It's obviously impossible to have every piece of steel end up in the same place. But it seem reasonable that a representative computer model building would fall down rapidly.
I'm sure it would be quite easy to get wrong though. I've had vast amount of experience of messing up simple physical modeling in games. Simple models introduce situations that don't exist in the physical world. It's a constant trade-off of accuracy against processor resources.