This paper proposes a modified Taylor-Kalman filter (TKF) for instantaneous phasor estimation based on an improved dynamic model describing the complex trajectory of dynamic phasors. The improved dynamic model is obtained thanks to a revised state transition equation of the rotating phasor and its derivatives. The proposed approach is assessed with dynamic compliance tests defined in the synchrophasor standard and in a power system test case. As the results show, the modified TKF can track the time-varying behavior of dynamic phasors under non-steady state conditions in power systems. It achieves good estimation performance without manually adjusting the Kalman gain as in the existing solution. Thus, the modified TKF fully exploits the self-adaptive nature of the Kalman filter principle.