The size-dependent separation of electrophoresing DNA chains of varying lengths has recently been demonstrated to occur in microfluidic obstacle arrays. The chain dynamics in the array may be modeled as a continuous-time random walk, wherein the motion of the chain in the array is interspersed with collisions with the obstacles, involving a size-dependent random waiting time necessitated by the chain unraveling and unhooking following each collision. Previous studies employing the continuous-time random walk model do not fully account for the electric-field dependence of chain extension during collisions in the array. In this study, we extend the continuous-time random walk model of chain dynamics with account for incomplete chain extension. We evaluate the accuracy of the model by performing Brownian dynamics simulations of DNA chains of different lengths in a self-assembled array of magnetic beads at various electric field strengths and lattice spacings and provide comparisons between the predictions of the model and simulation results for the chain mobilities, dispersivities, mean collision probabilities, and the separation resolution achievable between different chain sizes in the device. We demonstrate that the model correctly predicts the nonmonotonicity of the separation resolution with respect to the electric field strength.