Finding the current pose of the end-effector of a parallel robot is a problem, since its forward geometric model generally has several solutions. Current methods to address this problem operate mainly under the assumption that the robot never changes its assembly mode nor gets close to Type 2 singularities. Nonetheless, recent works proved that a parallel robot can change its assembly mode thanks to dedicated trajectory generation and control. Such a feature allows increasing the operational workspace of such manipulators. Hence tracking correctly the end-effector pose while crossing Type 2 singularities, is mandatory for a practical usage of this workspace enhancement method. However, on Type 2 singularities several solutions of the forward geometric model merge, making current tracking methods ineffective. To fill this gap, we propose a two-step pose-tracking methodology: First, a differential inclusion based on kinematics and dynamics is solved. Second, joint measurements are used to tighten resulting enclosures. The effectiveness of this method is discussed thanks to experimental data gathered on a planar parallel robot.