Neutron cross-section and their covariance at higher energies are generated by nuclear model parameters. The model based covariance exhibit a large systematic errors as compared to experimentally generated covariance. In the absence of probability distribution for the systematic errors, we propose the representativity factor approach to minimize these errors by simulating the upper and lower bounds by the Technique of Determinant Inequalities developed by us. We demonstrate the utility of our approach in simulating optical model covariance parameters for fluorine with reduced systematic errors.