Details re why IRV would not have chnaged result in Lizard vs Wizard race

Instant Runoff Voting (IRV)
presumably would have behaved in essentially the same bad way in this
election
because this race was dominated by the top 3 contenders.

Top-2-runoff and IRV are the same thing in a 3-candidate race, if the voters
are the same and act self-consistent.
But how solid was the top-3 domination in this race?
If 99% of Holloway's votes transferred to Roemer
while only 1% went to Duke that would have (just barely)
put Roemer in the runoff. However, a split that drastic would have been
extremely implausible. Trying for the most-plausible scenario:
If all candidates from Holloway (4th) onward were
lumped into a single "supercandidate" with

votes, then in order for them to put Roemer into runoff ahead of Duke, those 124127
would have had to transfer 80652+X to Roemer, X to Duke, Y to Edwards,
and Z failing to transfer to anybody,
for some X,Y,Z≥0 with 80652+2X+Y+Z=124127.
In practice
in the real world, even in simple IRV elections with
highly experienced voters,
at least about 10% fail to transfer, so I think it is safe to assume
Z≥12000.
I also think it is safe to assume Y≥15000.
Hence 80652+2X≥97127, hence 2X≤16475.
So in order to push Roemer into runoff we would have needed
a transfer from Supercandidate to Roemer of at least 88890
and from Supercandidate to Duke of at most 8237; this is a 92-8
percentage split. It seems safe to regard a split this drastic as having had
virtually zero chance of happening. This completes our argument
that IRV would not have changed the result.