In the past years, the emergent method called "electroencephalography (EEG) source connectivity"has gained increased interest due to its ability to identify large-scale brain networks with satisfactory spatio-temporal resolution. However, many related methodological questions remain unanswered and no consensus has been reached yet over a unified EEG source connectivity pipeline. The objective evaluation of the pipeline is challenged by the absence of a ground truth when dealing with real EEG data. In this paper, we show how a recently developed, large-scale, physiologically-grounded computational model, named COALIA, can provide such "ground-truth"models by generating cortical and scalp-level realistic simulations of brain activity. We investigated the effect of three factors involved in the "EEG source connectivity"pipeline: the number of EEG sensors, the solution of the inverse problem, and the functional connectivity measure, in the context of epileptiform activity. Results showed that increasing the number of electrodes (at least channels) leads to a higher accuracy of the reconstructed cortical networks, and that the weighted minimum norm estimate (wMNE) combined with the weighted phase lag index (wPLI) has the best performance at high electrode density. Although we believe that these results are context-specific, the model-based approach presented in this paper can be extended to address other methodological aspects of the EEG source connectivity pipeline in different contexts. We aim at presenting a proof-of-concept of the potential use of COALIA in the optimization the EEG source connectivity pipeline. © 2021 IEEE.