Lathe refrence frame shift?

Voltages other than +-2 v or more may be read by a voltmeter just because there is a small jitter/oscillation in position so that may not be significant. Its often easier to test to this with the servos disabled and just a slight manual rocking of the axis to get all the encoder states.

Also 2.5 is pretty old, is the Z axis problem new?

Unfortunately the encoder sample frequency option was added in 2.7. If its not a bad cable/drive/encoder and is a noise issue instead, its probably worth while to upgrade to 2.7.12 and try lowering the encoder sample frequency

Well I think it was that the encoder cable shield was not grounded properly. I had shield ground at the control side when I built everything. However I don't think the control had a proper ground itself. As soon as I connected the encoder shield to the servo drive ground the pulses became much more stable. Before my tenth digit on the control would fluctuate up and down, now it is dead steady. Also before sometimes when pulsing at .0001 it would take 2 clicks to get one digit, now it is 1 to 1 every time. I think its fixed, but I wont know for sure until the next time I have a long run of parts.