With decreasing inertia and short circuit ratios, the modern power grid is becoming more dynamic, increasing the need for dynamic state estimation (DSE). These DSE applications are dependent on data from the available data-sharing architectures, and are therefore challenged by the architectural limitations. This paper presents an analysis of how DSE is affected by these challenges, with a focus on incorrect topology and limited observability. Through a centralised Extended Kalman Filter (EKF) and Unscented Kalman filter (UKF) approach, DSE performance is evaluated using phasor measurement unit (PMU) and topology data, through a data-architecture abstraction. Both Kalman filters algorithms prove to be robust to temporary topology errors when estimating angles, while UKF outperforms EKF in terms of root-mean-square error (RMSE). Both are also capable of estimating complete phasors, angles and magnitudes, during high observability conditions. However, while phasor angles could be estimated during low observability conditions, the phasor magnitude could not be estimated by either approach.