One of the open challenges in transport modelling is to estimate within-day demand flows that reflect the complexity of individual activity-travel behaviour. While disaggregate (Activity-Based) demand models can recreate realistic daily mobility patterns at an individual level, they usually require an accurate knowledge of individual user behaviour (i.e. via travel surveys), which is not always available. As a result, practitioners often turn to aggregate demand models, that have the advantage of being less demanding in terms of data but typically under represent the demand for secondary activities. In this work, we take research on within-day demand modelling one step forward by proposing a framework that combines traditional methodologies with heterogeneous data sources in order to explicitly represent trip chaining at an aggregated level. We show that the combination of web-based crowd sensed data, network data and behavioural constraints allows to capture complex spatial and temporal correlations between demand patterns. The methodology is applied on the classical Gravity model to show how to incorporate within-day dynamics. Yet, any alternative demand model can be adopted. In our case, Generation and Attraction are used to estimate the systematic demand, that is enriched of information about individual activity patterns, and then a novel definition of impedance function based on Hagestraand ellipse theory plays a central role in spatially distributing locations of trips using geographic relationships and constraints deriving from space-time behaviour. A case study for Luxembourg City has been presented to show the potential of the methodology: the choice of using data from a different spatial context to account for the temporal dimension has been validated through comparisons with official statistics. The results of simulating a workplace relocation show the advantages of this new approach in representing demand related to secondary activities.