Do they have app-notes on this specifically? You say you're referring things back to the perspective of the PRM48 which makes sense but on the other hand the VTM48 is going to have a significant contribution to the overall transfer function. Are you accounting for it? Whether you need an additional factor of 24 depends on how you're choosing to model the entire picture.
But yes, as shown PRM48 errors are reduced by a factor of 24 (by the VTM) before being fed back, so I suspect your model needs to add this into account.
You're so big on LTSpice Treez, why not use that to simulate the bode plots of the control? Their transfer function was very easy to model in LT. If you posted a schematic of that it would be easy to see exactly what you're doing.