So now you have two sources of error. How do REALLY know your benchmark
car has an accurate speedo?

I immediately thought the same thing but on a reread of the post I think
he said he essentially got the same mileage with four vehicles.
(Probably the odd man is out.)

He is reporting about 6% error. I have checked car speedometers with
a GPS and finding them 2-3 MPH off at 50-60MPH is very common.
(and about the same error). It would be interesting to know the
tolerance on car odometers, but I suspect it not much better than that.

So I still don't think comparing the odometer readings is very useful
because you really need a known standard to calibate something.

The only way to do this right is to compare the obometer to a known
standard such as milemaker signs or a GPS.