I'm sorry, I still don't get it. Look at the diagram: B and R are continuously in contact with liquid mercury and therefore exposed to its (notoriously sky-high) vapor pressure. Entrainment be damned: I simply do not see how any combination of velocity and gravity could prevent that vapor pressure from remaining in equilibrium with the liquid mercury. Now if the working fluid were something with a negligible vapor pressure, I could understand it -- but, evidently, I would be wrong. What is going on?
R is not continuously in contact with liquid mercury. The removal rate from R is higher than the back-diffusion rate of mercury vapor from B.
You have said two different things. The second one is the key. R is in contact with liquid mercury, but that doesn't matter, because the vapor takes a finite time to propagate, and that is the trick. And that shows why the working fluid needs to be extremely dense. It is still counterintuitive; one would think the apparatus would lock up or implode.