Strange lost step problem, grounding related

Noticed today a real strange "lost step" problem on my mill. Machining
some SS shafts, I discovered to my dismay, that I was losing .025 in the
Z every time I ran through a fixture of 11 parts. Now, there are but 4 Z
moves in the entire program, so I was rather perplexed. I've always had
a niggling little creep in the Z, but within a range I could deal with.
This was too much. To make the long story short, I first ran a ground
wire from the inverter ground outside to a waterpipe that runs several
hundred feet underground. This knocked it down to about .003-.005 per
run, better but still a problem.
I have run many parts and multiparts, with peck drill cycles and minimal
or no Z creep, so I was looking for some variable that might account for
this. Discovered that if I run the inverter at 55hz, it really moves the
Z around. 60hz, no problem, or rather, back to the level of creep I can
deal with. Placing my hand on the graduated dial atop the quill ball
screw, I can feel it pulse now and then.

So, the inverter is bolted to the side of the motor on the mill and
presently grounded to a good earth ground, but the wiring is micky
(wanted to know if I was on the right track before doing it proper).
I have the cabinet with stepper drivers pretty well covered and
shielded, but wonder if I ought to attach a ground buss somewhere on the
mill and run all grounds to this, then run a proper ground wire outside.
Also, should I insulate the inverter from the mill itself?
All power and logic lines to the mill are several feet from any 220
lines, around the mill, but the extension cord that powers the computer
and power supply for the drivers runs alongside the 220/3ph line that
also runs down along the center of my shop (follows a structural beam).

Thoughts anyone? Further questions?

Thanks,

Jon