The calibration routine currently used, in essence, gets horizontal measurement dead-on at the expensive of the vertical measurement. I’m trying to develop a new calibration routine to overcome this… its taking time though.
This thread may help.