This study had the objective to develop an algorithm for accurate force decomposition in a redundant musculoskeletal system. The hypothesis was that the calculated load-sharing is dependent on the optimality criterion adopted, but also on the magnitude of carried load. The developed algorithm emphasizes that several established optimization techniques can be unified, by identifying and separating the underlying optimization functions and the numerical methods to solve the resulting system. A numerically efficient and easily adaptable solution method is thereby created. In addition, individual capacity values are introduced for the muscles, allowing the evaluation of a magnitude-dependent load-sharing, and a load carrying capacity of the whole system. By modularizing the optimization method, the algorithm can be used as part of larger simulation systems. To illustrate the possibilities of the algorithm, a model of the upper limb is used in a set of demonstrative examples. The results from the examples show how the interactions between synergistic muscles is predicted in different configurations, and at different load levels.