Quick answer, because simulating this kind of thing is incredibly difficult.
If you were to reassemble WTC 7 a hundred times, start exactly the same fires in it each time, and then wait for it to burn out, you would not get the same collapse each time.
Should be no surprise. Suppose you drop a bottle on the concrete. Sometimes it breaks. Sometimes it doesn't. Sometimes it just cracks, and sometimes it shatters completely. This is true even if you repeat the conditions as perfectly as you can. The collapse of a large, complex, damaged building is many millions of times more complicated.
This kind of system follows a science called
deterministic chaos. Even very small changes in behavior, early on, can lead to visible and seemingly large changes later. Say a weld hangs on for half a second longer than it did last time. Suppose these four bolts don't all let go at the same time. Suppose a column fragment just happens to get wedged between two large pieces and absorbs energy as it flexes, instead of just spearing through the debris.
You can see this effect clearly in the Purdue study of WTC 1 and WTC 2. They ran their simulations dozens of times, making tiny variations in the impact damage, the failure strain of materials, and so on, and found there are quite a few differences between runs. Nobody has bothered to do the same for WTC 7 that I know of, but it certainly could be done. I'd recommend anyone interested contact Dr. Irfanoglu at Purdue and start working on an NSF grant.
Now, having said that, the varying WTC 1 and WTC 2 results
do have some features in common -- like the fact that virtually all of them do lead to a collapse. That's what these simulations are for. They aren't intended to make things look exactly right. They're intended to provide some insight into the mechanisms of failure, and to determine whether a given hypothesis is
plausible.
And that's all NIST does with WTC 7. Their result did not depend on the exterior collapse simulation. Had this happened in the 1990's, the simulation itself would have been impossible! We wouldn't have simply thrown up our hands and said, "gee, we don't have a fast enough computer, guess we'll never know. Could have been fire, and could have been aliens." Indeed, you could remove that entire chapter of the WTC 7 report and lose almost nothing.
What they did, instead, was to set up two trials with different conditions, and see which one was a
better fit to what actually happened. This was a quick way to estimate whether the impact damage had any effect on the collapse. And the answer is, yes, it does -- although they also found it's not likely the collapse would have happened any sooner as a result of that damage. But neither case is a perfect match, nor does it need to be.
Truthers whining about how it "looks different" are following the
call to perfection logical fallacy. They arbitrarily decide how perfectly they want things to fit, and then complain when their threshold isn't met. You will note that there is
no professional complaint about the fidelity of the WTC 7 simulations. Those who understand the report, and why the sim was run in the first place, also understand that a perfect visual match is not expected nor necessary.
Truthers use "call to perfection" all the time -- show us aircraft debris with serial numbers, they say. Show us a video of the Pentagon impact. Show us the total mass of recovered debris near Shanksville. Well, you don't get to make those demands. If we applied them to the Truthers, they'd be silent forever. To wit: Show us your hypothesis. Show me that you aren't Paul Doherty. Show me that you actually believe the crazy things you're typing on the Internet. Can't do it, can you?