Backlash

I have been trying to chase down a backlash problem I have with my
Hiwin ballscrews for almost a year. I even sent the ballscrews back
to Hiwin and they said there is no detectable backlash in them. I
made a video of the test recommended by Hiwin. In the test I
attached a ball bearing to the end of the ball screw shaft so I am
not measuring any out of squareness in the end of the shaft. I am
using a tenths reading (.0001") indicator on the end of the ballscrew
and a thousands reading (.001") on the table. I am measuring
between .0003" and .0004" backlash in the end play of the radial
contact bearing, and .004" on the ballscrew/table movement. The gibs
are adjusted to that if I push as hard as I can anywhere on the table
I can see only .0005" deflection. The mill was recently reground and
hand scraped by a professional machine tool rebuilder.

There are only two places I can see the backlash coming from. One is
the yoke, which was supplied by Hiwin (very large cast iron, I can't
see that flexing), and the other is the ballnut/screw combination.
The ballscrew has a single nut. Hiwin says they take care of back
lash in one of two ways, either with a different pitch between the
nut and screw, or by adjusting the ball size in the nut. They won't
say which method is used in my installation.

Here is the video of the test:
http://www.flyingcritters.com/video/Backlash.wmv

Any suggestion would be greatly appreciated.

Vince