So got the new z-screws installed, after some reprinting of the z-screw parts. Didn't get the parts glued in at perfectly, there was a bit too large gap between the parts and motor doesn't have enough power to move through that without skipping steps. Newest parts work well as long as I don't print parts over 90mm in height.
Changing the screws was quite easy, luckily I designed the whole bed carriage to be detachable. Also the z-motor mounts are quite easy to detach.
https://www.dropbox.com/s/toqsp6n5ds...45.39.jpg?dl=0
https://www.dropbox.com/s/1t6tr260ce...15.10.jpg?dl=0

The new flex coupler works quite nicely, although the Filaflex tube is maybe even bit too flexible.
https://www.dropbox.com/s/gc0zejm6f9...15.21.jpg?dl=0
https://www.dropbox.com/s/81lchfvkyd...16.07.jpg?dl=0
https://www.dropbox.com/s/1fj4uf4rft...18.23.jpg?dl=0

Also printed part cooling fan using Taulman Nylon. Nylon had absorbed some moisture, which lead to stringing, but after cleaning the part it seems to work ok.
https://www.dropbox.com/s/t1tkh0p50l...06.26.jpg?dl=0
https://www.dropbox.com/s/vyxeiahun0...06.30.jpg?dl=0

Anyways, now the bed moves more or less straight. Another stackable box, green PETG. Again, having trouble with retraction, even though I upped the retraction length from 2.5 to 3.6. Starting to think that maybe the linear advance might be doing something nasty...
https://www.dropbox.com/s/s4vptji4l7...46.33.jpg?dl=0
https://www.dropbox.com/s/koi9dqru4p...46.40.jpg?dl=0
https://www.dropbox.com/s/7q3frm3hqb...46.52.jpg?dl=0

To fix the retraction problems, I ran KISSlicer destring wizard. To get a good baseline, I disabled linear advance for this. For string-free results, 5.3mm of retraction is needed. Next I'll run the same print, but this time I replace the increasing retraction length with increasing linear advance. Once the LA is tuned in, I'll run the retraction calibration again to get the hopefully optimal retraction settings...
https://www.dropbox.com/s/gz8as6q77e...06.13.jpg?dl=0

One thing Dollo has problems with is printing circles. I suspect that the friction of the axis mount causes very small stalls in the movement that lead to small imperfections. Loosening the motor mount screws should ease the movement, but I think this type of rail setup will never be perfect.
https://www.dropbox.com/s/vy2gh92k90...06.21.jpg?dl=0