I checked your video files (thanks for sharing them).
Current version of our software is not advanced enough to calculate the initial position of your camera #4 from your calibration video automatically. You have too few points and they all are too close to the room center, and calibration video is at 10 FPS which is too low for good synchronization. Also, there is a bug with camera adjustment tools (cannot move camera after +/-90 degrees azimuth) preventing manual configuration of initial camera position for camera #4.
Unfortunately, fixing the software will take about a week (the issue is very technically challenging).
I suggest the following workaround:
Please re-shoot calibration video at 30 frames per second using new iPi Recorder (version 1.0.1.36 Beta). Our recent research shows that our previous recommendation of recording calibration video at 10 FPS was wrong. Higher framerate results in better synchronization and much better calibration reliability and accuracy, even at 320x240. By the way, new Recorder monitors framerate and dropped frames on a per-camera basis, so it can help you diagnose potential synchronization problems (due to cables not plugged firmly etc.)
It is also important that in new calibration video you move pointer more randomly (like in circles and spirals) and cover bigger space. Imaging yourself a jedi practicing some fancy martial arts moves with a lightsaber in slow-mo. Irregular/random marker trajectories help to improve accuracy of calibration.
By the way, you may find the following video useful:
http://www.youtube.com/watch?v=8Pyly9BdcO4It demonstrates calibration of a more traditional marker-based OptiTrack system. While they have their own specifics (for example, they have high-framerate cameras so fast calibration wand movement is not a problem for their system), the video still demonstrates some common principles, like covering the entire capture volume and doing circular movements to make calibration patterns more random.