The moment frame connections at every floor kept the adjacent columns from bending.
The photographs of the area show that the adjoining columns did not bend.
The additional loads cannot 'skip' the adjacent columns.
NIST Apx. L pg 36
"If the initiating event was due to damage to the perimeter moment frame, then it would have started along the south or southwest facade."
"Analysis of the global structure indicates that the structure redistributed loads around the severed and damaged areas. Progression of column failure to
adjacent columns would have been arrested by the vierendeel of the perimeter moment frame which could span across a sizeable opening due to the strength and stiffness of the frame,"
The report then talks about possible damage to core columns and fires in the area of the
initiating event.
pg 37
" I2.1 South facade damaged
> I3.1 Perimeter moment frame redistributes loads around damage
> I4.1 Local failure only"
PG 41
"If a group of perimeter columns failed, the perimeter framing above this area would have redistributed its loads, due to the redundancy of the moment frame."
pg 42
" Initiating event scenarios from I4.4 to I4.6
> V1.1 Any perimeter column fails
> V2.1 Collapse does not progress vertically"
The report makes no further mention of the damage to the south face.
They site core damage and fires as the possible causes of the initiating event.
They ruled out perimeter column damage as contributing to the initiating event.
Only devout deniers still insist that there is a connection.
See ? This is what I don't understand about your argument. Not knowing anything about loads, I'd say that the weigth of the horizontal member would be distributed among the remaining columns, which means that ALL FOUR of the remaining columns, not only 1 and 3, would bear added load; though I suspect it wouldn't be an equal share, now that there is a big hole where 2 used to be.
The horizontal members went from column to column. They were not continuous.
The columns were continuous
If column 3 did not compress or bend, how could the load be transferred to column 4 or 5 ?