But does your PID display the setpoint or the actual measured value? If it is the former it seems you have a steady-state error typical of a P controller. So perhaps the I and D terms need tweaking?
EDIT: After reading your post again and reading a little about RIMS it realise I'm probably in...