On the validity of the effective mass approximation and the Luttinger k.p model in fully depleted SOI MOSFETs