It sounds like a mechanical problem. Something in the Z axis assembly is impeding the motion upwards that used to be smooth, so you're losing steps going up. This results in the DRO reporting a higher position than is actually achieved, since there's no feedback mechanism to tell it that it never got where it was supposed to. So then, when it goes down to the cutting height, the tool is lower than it's supposed to be.
To confirm (or disprove) this theory, run your Z axis up and down a few times, and measure the position of the tool at the top and bottom of the stroke. If I'm right, you'll notice that the measurements and the DROs disagree.