binary star - Peculiar orbit of circumbinary planet

I've created a Mathematica notebook file which should theoretically allow one to simulate any n amount of bodies. Whilst looking at a three-mass system where $m_1=m_2=2000m_3$, I noticed some peculiar orbital characteristics when the "planet" comes close to the "stars".
.

Pretty, but hardly what I expected. Are these kind of orbits viable, or is something going wrong here? Note in this case, the "stars" orbit each other in a near perfect circle.

For those with Mathematica, the notebook may be found here:

http://s000.tinyupload.com/index.php?file_id=07893360971974925836