This paper investigates the performance of a flow model in providing efficient travel-time estimation for varying flow patterns of freeway traffic by adopting a two-phase fundamental diagram. The model follows a discrete-packet-based mesoscopic simulation approach that explicitly considers both the anisotropic property of traffic flow in packet state updating and the uniform speed differentiation of vehicle packets at each discrete time step. The measure of travel time is obtained as a link performance resulting from a simplified dynamic network loading process. The spatiotemporal flow propagation on a selected freeway segment is simulated comparatively by incorporating both the proposed model and a linear-travel-time-function-based link performance model. Performance of the flow model in travel-time estimation is sought, considering actual measures obtained by a probe vehicle. The main improvement on estimating the travel-time process is that the employed model considers different speed and acceleration levels on different discrete time intervals and satisfies the anisotropy property by consistently simulating flow propagation within the dynamic network modeling frame. In contrast to the vast data need and computational burden of trajectory-based methods, the employed flow-based model requires only the time-varying inflow profiles to estimate spatially and temporally varying travel times by artificially segmenting freeway routes.