Most Fleet Assignment Problem (FAP) formulations use a leg-based estimation of revenue loss to derive the passenger revenue component of their objective function. This neglects the leg interdependency of revenues, caused by multileg itineraries. We tackle this problem by modifying the objective function using information provided by a passenger flow model devised by two of the authors and presented in a previous paper. It models spill and recapture between itineraries, accounts for the leg interdependency of revenues and does not control passenger flow to the airline company's advantage. We iteratively improve the objective function of the FAP by alternately generating fleet assignments and analyzing them with a modified version of the passenger flow model. We have tested this process on a large-scale network made up of Air Canada data with various demand levels and distributions. The expected profit improvement is substantial, most of it is reached in the first few iterations, and the objective function adjustment takes on average less than half the FAP resolution time.