Yes to extremely fine feeds --
one spark or two is enough. 
Out of roundness, Dremel runout, and the flexibility involved here are not issues for precision, assuming you don't try to break the grinding rig by bouncing it off the jaws with a heavy feed.
Here's why:
1.) The stone is rotating. The jaws are rotating. The stone is moved axially
all the way through the jaws parallel with the ways (because it is driven along the ways by the carriage). Therefore the full length of each jaw is subjected to the same high spots on the stone and the same run out and any other errors, and these are applied to all jaws along their full length, parallel to the ways. All of the jaws see the same thing, parallel to the ways. The only way a taper could be introduced, is if the headstock spindle itself is not parallel to the ways.
2.) With extremely light pressure (which is ALL you want, assuming a reasonably true chuck-- a few individual sparks with each feed adjustment) you are not wearing down the wheel at all. Any high spot on the wheel is just ticking a high spot on the jaw, removing metal. When that stops, and not before, you can feed the stone outward a very tiny amount again with the cross slide. Then make passes again through the bore with the carriage at that position until the ticks stop again. Check that all jaws show some shine along their gripping edge. Done.
3.) Bending is not an issue with very light feed, and waiting for the ticks to stop. When the ticks stop there is no bending. There is no force on the stone. Just before that there is infinitesimal force on the stone. Think of it as reaching the limit of an equation. Actually, bounce, shouldn't be there in the first place -- this isn't hogging on a grinder. It needs to be done slowly with a light touch. Any supposed bounce will reduce to nothing at the point when the jaws are no longer grinding at all.
4.) The high jaws are reducing in size as you grind them. Eventually after a few rotations and feeds all three are the same height. Nobody is bouncing or even grinding when the ticks disappear. Each jaw is subjected to the same conditions, and the all level out. That's the principle.
I found the reference URL that I used to do this. It worked perfectly for me with the Dremel.
By running the chuck at a reasonable speed, the jaws were tight to the scroll by centrifugal force, without using the usual spacers or rings. Any flexibility or run out in the Dremel was cancelled out by the method of very light feed, waiting until the ticks stopped, and passing the stone all the way through the bore. It worked. Not theoretical. Just actual. Here's the reference to the original:
http://www.varmintal.com/alath.htm (point of current interest about 1/3 down the page)
