When you get to tenths thou and similar level errors process becomes almost more important than dead-nuts equipment accuracy. So long as your vice or other tooling is decently good the closeness to "perfection" of the job or set-up depends on how you approach things and how well you do the job.
I tend to the view that most of then time your true grip reference is the fixed vice jaw. So long a that is true and vertical a decent set up can be achieved even if the vice base isn't dead flat relative to the mill table or exactly perpendicular to the vice jaw. So long as the component is sturdy enough to grip without distortion something can usually be worked out by appropriate selection and positioning of parallels along with use of resilient packing or a roller for line contact on the moving jaw.
Generally I favour resilient packing, eg soft aluminium, as being easier to handle than a roller.
One day I shall experiment with magnets and a D shape packing piece set flat side to the job to combine the flexibility of a roller with a larger contact area on the workpiece..
My normal use vices have a very slight perpendicularity error between the fixed jaw and the base. Just enough that if 1/4" wide parallels are placed next to the fixed and moving jaws with a robust test piece having truly parallel sides around 2" or more wide gripped between the jaws and beaten down the parallel next to the moving jaw is just loose enough to slide out with moderate pressure. Depending on the job I either rely on the parallel next to the fixed jaw to set position, but still fit one close to the moving jaw to ensure things don't tilt objectionably when beating down, or put two parallel at the 1/3d jaw separation points. Seems to work adequately as parts look too be standing up straight.
Clive