In CALPHAD, parameters associated with lattice stability and excess energy are refined against empirical data and used to construct descriptions guiding phase equilibria and thermochemistry. System complexity and size would limit database creation as the number of optimizing parameters may become extremely large. In this work, potential simplification of CALPHAD optimization was explored with a system exhibiting four terminal solutions and four intermediate phases. Upon least squares optimization, final values presented reflection to and were constrained by starting values, forming a cloud of candidate minima for each parameter. This implies most final values that one finds during refinement may not be global minima. While each local minimum showed the smallest error sum reachable per starting value, if parameters are not considered as pairs (i.e., a + bT, not a and b independently), final values would be limited to only local minima, minimizing the chances of attaining a global minimum.