Most of the current formalisms for flexible multibody simulation lead to very computationally expensive models. Some applications, such as Hardware-in the-Loop (HiL) require models which can be evaluated at very high rate or even real-time. In order to meet the need for highly efficient flexible multibody simulation, Brüls introduced a system level model reduction technique, namely the Global Modal Parameterization GMP. This method generates very computationally efficient models but is less suited for mechanisms with a large number or rigid degrees-of-freedom (DOFs), since the required storage space to store the reduced model rises exponentially with the number of rigid DOFs. In order to resolve this limitation, the author proposes an adaptation of the original GMP method, namely the Sub-System Global Modal Parameterization (SS-GMP). In this approach, GMP is only used to reduce the coupled DOFs, which are expressed in a frame of reference connected to the mechanism, and the global motion of the system is described by the motion of a system attached reference frame. The paper discusses how the reduced equations of motion can be derived for the SS-GMP method based on a Lagrangian approach. The exact formulation of the equations of motion is dependent on the description of the original model, and the method is validated against an unreduced model described by a nonlinear finite element approach for flexible multibody simulation. The potential of the method is shown through a numerical example of a quarter car model. This example demonstrates how the SS-GMP approach can strongly reduce the computational load for the simulation of a practical system, while still providing accurate results.