I'll throw a flyer out there. The incremental onset of friction force vs normal force determines if the cover milks. If your pinch style device gets both force going at about the same time there's probably little milking. If a device pinches the $@!% out of the rope and applies no tension there's zero milking. If you put on blue latex smurf gloves that grip with almost zero normal force and apply tension, well that's about the best way to milk a rope cover. In a rope bending device the bend is under tension and the bollard gives simultaneous normal and contact friction forces. They're probably divvied up pretty evenly hence little milking. The normal force transmits tension to the core (via friction) kind of like the cover is an extension of the bollard surface.
I'm sure the voodoo of balancing settings for rope "braiding" twisting, etc during construction is tradesman, machinery, fiber etc specific and pretty much trade secret. Think about the chinese finger trap action, elongation of such, elastic fiber elongation (core and cover), fluffy lay till tensioned and putting that through a bend or knot under tension and keeping it all good. Empirical is a word that comes to mind. Rope manufacturers even somewhat brag about achieving just right cover tightness and so on.
I once toured a Canadian plant and the fellow made a few comments about the machinery as we went by. Spools of yarn everywhere. Quite impressive. He pointed out the tensioning devices.
Near ultimate strength load sharing is probably beyond chinese finger trap forces. That's a reason to configure rope terminations and allow sufficient length under tension for core and cover load equalization, well, not equal but distributed according to design intention. Splice technique could imbalance core and cover. With a long rope the imbalance distributes to become a smaller percentage deviation. On a short foot or two segment the percentage is amplified. Then one or the other portion becomes overloaded before the other reaches full load.
Cold, windy, "cabin enclosed".
best to all