To design processes that effectively use polymer directed self-assembly, we would like to have a complete picture of stable and defective polymer configurations. Field-theoretic simulations are an effective way to gain knowledge about these configurations and predict defect populations: we can easily vary design parameters such as prepattern dimensions, wetting conditions and polymer composition/architecture and observe their effects on pattern formation. We previously showed that an optimized phase field model, a modification of the Ohta-Kawasaki model, is more accurate at predicting domain spacing and defect formation in bulk systems. This accuracy is achieved by a systematic mapping procedure that optimizes parameters in the model using inexpensive, low-dimensional selfconsistent field theory (SCFT) calculations. We now make two improvements to the model. First, we implement a conjugate gradient method, a more efficient numerical solver for phase field models not available to SCFT, and characterize its performance. We find that an optimized phase field model simulation requires one to two orders of magnitude less computation time than an SCFT simulation of the same system. Second, we extend our model to confined templates and demonstrate that the new model does not suffer from the nonphysical behavior found in other phase field models in the presence of confining walls.