We present new correlations revealed by our study of polymer mixture miscibility. Applying our simple lattice-based equation of state, we search for patterns in a large sample of experimental blends and uncover some intriguing relationships. One such correlation connects the value of the difference in pure component energetic parameters with that of the mixed segment interactions, suggesting new possibilities for predictive modeling that would require only pure component data. Our work reveals different patterns for UCST-type and LCST-type blends which we connect with physical underpinnings reflected in the microscopic parameters. Throughout we emphasize the importance of modeling protocol, most notably, the importance of carefully and consistently applying a fitting procedure to the same temperature range for the two pure components. The reasons for this are demonstrated and discussed in terms of a careful analysis of the model behavior which includes a look at the sensitivity to pure component fitting. We provide details (applicable to other equations of state as well) on the methodology needed in order to obtain the most robust and consistent results.