That is a good question DB.
I have argued this issue before, to no avail. IMO the currents flowing on these radials are NOT CM currents, as I think you suggest.
In my S4 model's tabular currents report it shows the minus (-) current magnitude on the radials noted to be a fraction of an amperage difference greater than the (+) current magnitude on the 1/4 wave section of the monopole radiator in the middle. Pretty simple idea for cancellation.
As the model also shows these two currents cancel almost completely leaving a small (-) current flowing into the far field. This current is constructive with the top 1/2 wave radiator, and in my opinion this is where the non-apparent collinear idea came from. I simply claim that this current contribution is miminal at best.
IMO, the group consideration here justifies their idea based on the cone being coaxial. Like you suggest the cone is arguably not a feed line.
There might be CMC's flowing on this antenna from the termination point of the coax (the feed point), but they will be flowing down the coax...not on the cone.
CMC's are described to flow in every other situation, where we have currents flowing down the feed line from the unbalanced feed point, instead of flowing on the attached radials as we would prefer.
