Symbolic manipulation uncovers hidden constraints for a model and facilitates model reduction for solution. The present work points out possible pitfalls of this procedure and finds possible solutions. The performances of different reduced forms of a condenser model are compared under step and impulse perturbations. The system does not need to be at steady state before perturbation. Some symbolically manipulated models cause inconsistent reinitialization and convergence and introduce significant computational errors. Physically unreasonable system behavior thus emerges, which worsens with increasing model complexity. Nevertheless, the less symbolically manipulated models present realistic and accurate behaviors. Another approach to model reduction is the reformulation of a model through a physical insight. A physically reformulated model proposed for the system leads to accurate and physically reasonable system behavior.