Meh. I hate proofreading other people's spreadsheets. However...
You've assumed a linear increase in speed for each stage of the fall, which isn't quite right but not too far off. I set the energy loss to zero and got just over 12 seconds for collapse, which looks about right for a floating floor model with losses. Your collapse time is very much more sensitive to collapse energy because of the increase in energy by storey, but that's what would be expected because that's how the columns were actually constructed. Overall this looks reasonable, although I may have missed something somewhere - I haven't dissected it fully.
What does make a huge difference to the predicted collapse time, as I commented in another thread, is the actual failure mechanism, specifically the number of buckles per storey. Changing from three buckles per storey to one per three storeys - which, based on the column lengths in the debris, appears closer to what actually happened - makes a difference of about a second for every ten floors in the early stages of the collapse; for example, if you assume the first twenty storeys failed one at a time and that the collapse progressed by three-storey collapses from then onwards, the collapse time comes down to about 15.3 seconds. While I take your point about not over-refining a very simple model, this is relevant to the 1-D approach, so I think it's a valid line of discussion.
What I think this shows is that the collapse time is maybe more dependent than I initially thought on the detailed collapse mechanism.
Dave