Rolled vs. Ground ballscrews

Alright guys, maybe some of you are open to a little debate.

I am in the design phase (still) of a 5-axis mill and have purchased
nearly all of the pricy stuff except ballscrews. Now, obviously
ground ballscrews are more expensive due to their higher accuracy,
etc.

My question is, exactly what are you getting for your money and is it
posssible to achieve the same end goal using rolled screws and a
little clever design?

If the major difference is the accuracy of the lead and backlash,
then wouldn't it be possible to eliminate the backlash in a rolled
screw either by using a double nut design (preferred) or repacking
the ball nut with larger diameter balls? Then, once the backlash is
minimized, the only real concern is the accuracy of the lead.

Now, my thoughts here are that this is only a concern if you are
using the servo motor's encoder to close the feedback loop by
interpolating the axis position from the number of rotations and the
diameter of the screw. However, if you used linear encoders on the
axis (preferred) or used an arrangement with a timing belt fixed to
the slide, turning a pulley which drives a rotary encoder, wouldn't
that give a very precise readout of *actual* axis location? Provided
there is *no* backlash in the ballscrew arrangement, there should be
no problem with hunting on the servo mechanism, etc.

As I have said before, I base my statements on absolutely zero
experience and no intelligent thought whatsoever. Just an idea
I had.

What do you guys think? Any problems? Better ways?

Jim