Hi,
If there is an uncontrolled part of the system then that part is not being properly represented by the feedback, assuming it is controllable in the first place.
For example, if the oscillation is near 1kHz and the output filter (270uH and 100uf) attenuates around 1kHz, then the oscillation may not reach the output well enough to influence the feedback in order to get the control to reduce it's effect on the output. In other words, if the only feedback is from the output and the uncontrolled signal can not reach the output, then the feedback can never compensate for that error.
One solution might be to get the output filter cutoff frequency farther away from the bad signal frequency, or perhaps modifying the feedback network to allow more of the problem signal through.
Another solution might be to modify the feedback network to include feedback from the actual problem signal more directly, although this is probably harder to implement in hardware.
Actually the right way to do this is to do an analysis of the whole system, by replacing the control chip with a linear version and doing a stability analysis. This takes a little work on your part but would provide you with more insight about what works and what doesnt. For example, how did you choose the output filter values...and if you read them off of a schematic designed by someone else, how do you know they are a good choice unless you do an analysis.
Another question is how much this oscillation affects the system as a whole. Does it hurt anything such as the output or the efficiency.