Hello Marzio,
Yes, that is feasible, but it is true that modeling an MMC of this size would be much easier to do the existing way using the Power Module blocks and their inbuilt vectorization capability. Do you think it’s accurate to get the loss and junction temperature information for just a single cell and extrapolate the losses for all cells/legs from there? If so, you could place a pair of standard IGBTs and Diodes in the series stack and then use n-1 for the remaining cells in the Power Module.
Regarding the thermal parameters, I think you should contact the manufacturer for recommendations.
Kris