This paper considers nonlinear predictive control via embedding strategies. A novel robust MPC algorithm for input-saturated uncertain discrete-time linear systems subject to norm-bounded (LFR) model uncertainties is presented and contrasted with direct nonlinear and robust polytopic MPC methods in two final nonlinear examples: a mildly nonlinear two-tanks hydraulic system and a highly nonlinear CSTR plant. From the examples it results that the embedding strategies compare favourably with direct nonlinear MPC in the mildly nonlinear system whereas their performance may degrade, often considerably though exceptions are not rare, in highly nonlinear plants. However, the numerical savings achievable by adopting embedding approaches are remarkable, especially for the proposed solution.