Hi Alex,
Your last statement gives a clue as to where some of the error lies. The loading of the belts on the U3 is quite high and this is taking up some of the play in the spindle bearings.
I would also want to check that the Quill is parallel to the column. Shimming the Column you say has little or no effect. This I would expect and leads me to think the Quill is pointing either towards, or away from the Column, (you did not say where the largest error was in this plane)
Resting the Column, and Mill Head, less motor on your larger mill table. Place vee blocks under the column if you have them or two good parallels and place clamps over the Column, but do not tighten. Adjust the mill head to be parallel to the table. Juggling the sizes of the parallels to be the difference in radii of the Column and Quill will make this a quick and easy set-up. Then clock the Column to be parallel with the X-axis and clamp. Move over and clock the Quill above and below the milling head. The Quill will need to be clamped and you will need to raise and lower your Z-axis. One, to get the high spot over the Quill and two, to clear the mill head casting when you traverse the X-axis.
Don't be tempted to clock a Silver steel bar in the collet or drill chuck for the above test. This will introduce further errors due to the type of work holding used.
While at this setting put a clock on the Spindle, at the Spindle nose and wiggle it to and fro. This will tell you what play is present. From what I remember this spindle runs in one long Oilite bush.
Regards
Gray,