The closest I could suggest is 'contact bounce'. The surface condition (ie oxidation) could also come into play. Also, there could be microscopic bits of dirt and dust that could help mess things up.
While you may think you are in constant contact with the iron rod, the fact that it has a rough surface means there will be hills and dales which you will be bumping into and skipping over, respectively. As a result you will have times when the sliding contact is, in fact, not touching the rod. It will be at a microscopic level, but at low voltages, that's all you need to interrupt current flow.
Try the same experiment, but with a machined, polished metal rod and you will see much less of the effect. (Even a polished surface is rough at microscopic scale.)
To reduce the effect, you could try a capacitor across the LED to store some charge.