Looks like you've managed to squeeze a fair bit out of the trailed data. I agree with atalas, if you can get some guiding involved (assuming you're not already, in which case it could be differential flexure) more detail will come out when you get nice round stars
