Extended Tracking extracts feature points from the environment dynamically using frames from the device camera. The more feature rich the environment is, the better the tracking experience.

Reviewing your video, the environment is a whiteboard with some black writings. This is a very challenging environment for the tracker optimally perform.

Please see this article for the feature overview: https://developer.vuforia.com/library//articles/Training/Extended-Tracking

"The more detailed and feature-­rich the environment, the better Extended Tracking works."

Vuforia 5.5 can support an 'Extended Tracking' like experience when the enviroment has limited feature points. Suggestion is to use a combination of Image Targets (gets the initial pose of the target and thus define the plane) and Rotational Device Tracker feature (to update the pose without having to extract feature points as required by Extended Tracking).

https://developer.vuforia.com/library//articles/Solution/Using-the-Device-Tracker