Thanks a lot, you really fixed the offset! I'm not sure how, but you did it.
I also increased the repeat remapper of the gradient to make the image keep its square dimensions.
However, your bias for the consecutive rings that you put in is not very accurate. You're right, I need a bias, but not at the parameters you set.
At a repeat of 2, each consecutive ring should have exactly 1/2 (1/repeat) the radius as the one it is inside. If you remap the altitude of the bias with an 8-step (there's 8 boxes each way on curve components) stairs curve (threshold at 0), then each consecutive vertical row should be 1/2 as filled as the one below it.
I admit I don't understand how the bias curve works.

So I'll be toiling with that...
I believe my assumptions are correct, but please save me the pain and tell me if all the repeat stuff I explained has nothing to do with circular tunnels.