Nikkor 20mm f/3.5 Ai (said to be pretty good at higher enlargements, thanks Damil)
Zeiss Flektogon 20mm f/2.8 (former Eastern Germany lens)
Nikkor AF 50mm f1.8 (cheap standard lens from my 1:1 shootout)
Schneider Componon-S 50mm f/2.8 enlarger lens
CnScope 4x ND 0.10 $25 microscope objective
Another Chinese 4x ND 0.10 objective from British Ebay seller
Leitz 4x ND 0.12 lent by forum member Emsiem (thanks Neno!)
Noname Chinese 10x ND 0.25 set to lowest magnification

At the back row you can see relay and extension devices; Nikkor 85mm f/2 Ai, 3 stage extension rings, Nikkor 200mm f/4 Ai and 150 mm bellows.
All shot using Nikon D7100 and Godox ML-150 macro ring flash in the dark with 2 sec exposure and second curtain sync, all the lenses reversed. Butterfly wing shot as 11 segment stack at 40 micrometer steps. All the lenses/objective were properly shaded and extensions/adapters covered in black matte finish.
In order to get as close as possible to 4x magnification, extensions were adjusted resulting by the following working distances (all measures in mm):

My first concern was the way I should connect the lens to the camera. I'm not sure I have ever read whether it's better via bellows or relay lens. So I have done it both ways. I believe I have found the answer once and for all (at least for those particular 3 lenses at 4x magnification):

Upper row using 85mm and 200mm lenses, lower with bellows/extension rings. Contrast difference is huge and disqualified extension way from the remaining tests.
Now we get to the main part. Here are center crops compared. No big differences actually, Leitz objective leading, with the Nikkor 20mm as the first runner up:

Important detail to point out is the optimal aperture: Componon and Nikkor 20mm did their best at f/4 while Nikkor 50mm and Zeiss peaked at f/5.6. Although I did my best to keep flash to object distance constant, slight variations resulted in different shades and detail presentation.
Differences extremely escalated when moving to the picture edges so at the corners it looks like this:

Again, differences are not huge IF the lens/objective is designed for this kind of image circle size. Three lenses performed terribly in this test: 10x ND 0.25 mounted directly to camera mount (approx. 4.5x magnification), Another 4x Chinese Objective and Leitz 4x 0.12. The first one was not expected to do any good since it was pushed way out it˙s design limits while Leitz is actually not an achromat,, requiring eyepiece compensation.
Your comments are very welcome!