Public transport systems create large numbers of possible combinations of public and private modes that travellers can choose to use. As a consequence, current travel demand models face problems related to flexibility in the modelling of how travellers value and use modes, consistency in the choice process and computational efficiency. To resolve these problems, on theoretical grounds, a supernetwork approach is found necessary, where a hybrid form of the network GEV model and the path size logit model should be used to model the multi-modal route choice. A case study for mode and route choice in Île-de-France shows that this proposed approach is feasible in practice and confirms that the constructed flexible modelling framework indeed possesses most of its suspected advantages over current travel demand models.