One of the main oscillation physics goals of the Deep Underground Neutrino Experiment (DUNE) is the measurement of CP violation. To achieve this, DUNE plans to employ liquid-argon time-projection chamber technology to precisely capture neutrino interactions which require a highly sophisticated automated reconstruction software such as Pandora to fully characterise. Pandora uses a multi-algorithm approach to gradually reconstruct the structures and features of neutrino interactions; having been deployed successfully for the MicroBooNE and ProtoDUNE experiments it is now being optimised for the DUNE far detector. This poster will illustrate how the multi-algorithm approach can be exploited to optimise the sensitivity to the CP violating phase at DUNE. An iterative approach has been used to identify the leading challenges in the reconstruction for this important analysis and has motivated new Pandora algorithms to improve the reconstruction of shower-like deposits, which are key to reconstructing electron neutrino interactions when searching for CP violation. The potential for large gains in sensitivity is demonstrated.