I need to double check but I believe what is happening is: you are executing the STOP instruction, the clock is stopping, the PLL is (as expected) losing lock, and the resulting interrupt is waking the part up again. In this case the solution would be to disable the PLL interrupt before executing STOP.
When entering WAIT mode you have the choice of deselecting the PLL automatically if required - see the PLLWAI bit for more info. Although, in that mode it may be useful to know if the PLL loses lock.
You need to be careful when entering STOP (or Pseudo-STOP or WAIT) that all of the possible routes to waking you up again are not pending otherwise you will bounce right back again.