Large scale outages on critical infrastructures (CIs) resulting from small initial failures, although infrequent, are increasingly disastrous to society; thus, it's imperative to optimally design these systems in order to maximize their resilience against cascading failure. The optimization task requires the understanding and modelling of the dynamics of these systems. In this paper, we consider two approaches to infrastructure modelling: (i) a simplified (computationally cheap) graph-theoretical (topological) approach based on network theory; (ii) a realistic (computationally intensive) physical approach based on power flow models. The objective of the work is to investigate whether the optimal network obtained using a simplified topological model is also physically optimal when a more realistic power flow model is embraced. This is performed through simulations on the 400kV French power transmission network. A discussion of the usefulness of topological models to study the dynamics of cascading failures in electrical infrastructures is also contributed.