Disaggregated models, such as activity-based and random utility-based frameworks, play a central role in travel behaviour analysis and urban transport planning. However, conventional modelling practices often follow a sequential estimation-calibration process that introduces challenges such as error propagation, inconsistent value-of-time estimation, and poor alignment between micro-level behavioural outputs and macro-level system performance. This paper addresses these issues by proposing an integrated modelling framework that simultaneously estimates discrete choice parameters and calibrates system-level constraints, such as observed traffic counts, OD flows, and reference Value of Time (VoT) targets, within a unified optimisation structure. The proposed approach embeds macro-level calibration objectives directly into the estimation of mode, destination, and route choice models, enabling coherent behavioural interpretation while ensuring system-wide consistency. We implement and evaluate this framework on synthetic networks with varying complexity, employing both global optimisation and Bayesian calibration methods. The results demonstrate that the developed model variants consistently outperform traditional log-likelihood-based models in replicating key system metrics, while maintaining plausible and stable behavioural parameters.