[Christ]

The first thing I notice about those images (I haven't tried to translate it yet) is that the cylinder is approx 25% the size of the cube, and placed the same distance from the cube as the width of the cube.

Scaling this up (scaled models show basic fluid dynamics, but do not catch the nuances), the cylinder would then have to be scaled, the cube (trailing object) gets scaled up, and so does the distance between them.

There is also (intuitively, I admit) an inversely scalar difference between leading object size and distance from the trailing object. The larger the leading object is, the closer it can be.

Based on those images, I think we're looking at virtually projecting a 10-12* angle from the absolute edges of the trailing object forward, then placing a cylinder at the convergence.

EDIT: Adobe and FoxIt both view the entire file as a picture, so I can't select individual text and xlate it. I'd need OCI and a Japanese text file to do it.