We develop methods for alleviating the major impediment in the extension to larger and more complex systems of our matrix method theory for describing the long time dynamics of flexible polymers and proteins in solution. This impediment is associated with the enormous growth in size of the required basis set with the addition of higher order mode coupling basis functions, which an needed to describe the influence on the dynamics of the "internal friction," or equivalently of the memory function matrices. We use the first order eigenfunctions (the generalized Rouse modes) to construct an approximate mode coupling basis. Specific applications are made to united atom models of alkanes with a white noise structureless solvent, where the theory is compared with Brownian dynamics simulations to provide a no-parameter stringent test of the theory. Good convergence is found to the full second order treatment with the new basis set whose size scales more nearly with the size of the system rather than the cube of the system with the previous full basis. These technical improvements enable us to test the need for third order contributions to the dynamics of the longer alkanes and to compute the orientational time correlation functions probed by fluorescence depolarization and NMR experiments. Additional symmetry considerations provide further reductions in the required basis set sizes.