There's no doubt that you'll get a more consistent behavior from a more consistent contact profile. But I don't get the impression that that's the problem here. If it were the problem, then the solution would be to increase the tension between the top and bottom bearings, which was the one prong of the original post. (There's also plenty of anecdotal evidence on this forum that over time the steel "V groove bearings" will eventually conform the comparatively soft aluminum rails to such a profile anyway.) The other prong was removing axial slop from the v-groove bearings. 

I don't think the goal is to "eliminate all axial loading" I think the goal is to eliminate all slop in the axial direction without putting axial loads on parts that aren't meant to accept axial loads. I think that using a modest axial preload on the outer race of the v-groove bearings to eliminate that slop is an reasonable approach, given that V-groove bearings are ostensibly designed to bear axial loads.

The trick seems to be finding the right flanged sleeve bearings. I didn't find anything on McMaster available in the sizes we'd need for this particular job in 5-10 mins of searching. While the quasi-floating shims illustrated in the original proposal can do the job of coupling the outer races of the bearings, I would prefer a more stable, less float-y solution, like the one illustrated in my quick-n-dirty illustration here.

The summary of my point was that you supported a strong axial bearing (v-groove), with a weak axial bearing (radial bearing), but if you supported it with another strong axial fag bearing (thrust bearing) it would be an improvement.

1.) The v-groove bearings ARE rated for axial loads (as well as the obvious radial loads), yet they are deflecting when applying force on the axial axis "lever". They should be able to handle that, but they aren't.

2.) Your solution is to compress/sandwich the v-groove bearings, between one/two other "things". This provides both a wider stance on the axial bolt, and some preload forces.

3.) However, you chose to use regular radial bearings. Unlike the v-groove bearings, or thrust bearings, regular radial bearings are almost never rated for axial loads - they are only meant to take radial loads. Some radial bearings will actually even have slop axially.

4.) Using thrust bearings (I said "washers" incorrectly, before), or another item designed to take axial loads, will likely give improved results. They will be intended to handle loads in this direction, so they will experience less deflection, and will live longer.

5.) You still may need to shim them. All that your shimming is doing is mechanically connecting the two koyo bearings, because otherwise there is a gap.