Don Foreman writes:

This wouldn't hurt function as long as you take up that slack in use ...

You might think that. But it does seem to be a problem, because when the
indicator probe is offset from the spindle axis, this adds a component to
the indication that is not aligned with the X or Y direction, making it
nearly impossible to zero in on the sweet spot. You are hunting around a
two-dimensional space for a null, but without a monotonic slope to follow.

Essentially you are trying to zero out two different, independently (mis)
aligned offsets at the same time, the spindle to the indicator, versus the
spindle to the work, and the axes of table motion affect each differently.
So no matter how I turn the dial body or twiddle the leadscrews, the last
few thou won't disappear. You zero one axis, and the other unzeroes.