Correct me if I'm wrong (I deleted my copies of the NIST reports a while back), but I seem to recall that NIST didn't force their FEA calculations to stop after initiation. The FEAs eventually broke down due to increasing structural failures going beyond what the software could handle.
My understanding is that we do not have the capability to come anywhere close to reliably modeling a collapse progression like what occurred in the towers or in building 7. If this is correct, it is not that NIST neglected to investigate how the buildings collapsed even though they could have done so. It's that they recognized that they could never do better than vague rough estimates (and that modeling progression was unnecessary). In other words, the investigation was as compete as was feasible.
Can anyone confirm whether this is correct?
It is correct both for the set of reasons you identify and for one other set of reasons which are often more significant.
They come from opposite ends of the spectrum of approaches to "solving the problem".
I will deliberately oversimplify to explain the point.
One end is going for more and more detail which faces two big barriers:
(a) It has historically often overloaded the available resource to do the complicated (These days FEA) calculations; AND
(b) The reality that the starting point data is not available at the level of deeper and deeper digging into detail/granulation.
Most modern generation engineers identify with FEA which has some advantages (big ones) but a few built in traps. One of them is losing sight of whether or not you have the detail.
The other "end" of the process is even more fundamental - "Is the problem properly defined?" AND "Do we have a valid starting point scenario?"
Wrong "starting point" can lead to lots of going round in circles - two of the most obvious ones on forums such as this being T Szamboti's "Missing Jolt" and the generic version of the same underlying error - "would tilt of the "Top Block" cause or prevent axial impact of dropping column ends"?
Both those are big picture errors of false starting point. The presumed and accepted by both "sides" starting scenario never existed. Both sides making or accepting a false starting point. No amount of maths or FEA wil get you out of that trap.
And there are many smaller but similar issues which arise at more detailed levels - when the starting data does not support the presumed accuracy.
So that is a simple overview - I agree with your specific points.
Have another look at your "they could never do better than
vague rough estimates". Because for that one - progression - rough estimates were not "rough" at the level of detail required to solve the problem. Yes they were "rough" for the often presumed objective of putting multiple decimal points into the numbers. "Number of decimal places" is rarely the goal. And accuracy for the real goal only needs to be "sufficient for purpose or better".
A similar issue from a different perspective. I have many times posted two assertions viz
1) the rapid cascade process of Twin Towers "initiation" was led off by failure of one column for which heat was the trigger; AND
2) We will never know which column it was and we do not need to know to prove cascade mechanism.
That second assertion is not acceptable to some members - they are from the "need all the details" mind set.