I think that the only factor which determines the angle of a cone what the dish "sees" - is its focal length.
This was my initial suggestion submitted in my original post. I asked for other Members' opinion, and they stated that the dish's selectivity depends on its diameter: the formula (equation) in Wikipedia for "angle of view" = theta, contains no focal length as a factor. I disagreed, but recently realized that I was only partially right:
The angle of view - indeed - depends ALSO on the dish's diameter, but - not exclusively: it depends on its focal length, too.
For explanation, the reverse process could be used : the LNB could be assumed as a source or radiation so the dish could truly ILLUMINATE the Clarke Belt, as if it were driven by B.U.C.
Using some idealization, first we can imagine that in the focal spot there is ideally small source of radiation, sort of geometrical, dimensionless point. If the paraboloid is also ideally correct, then the radiation reflected from dish would take shape of cylindrical beam, going to infinity, all rays - parallel.
In reality, there is no dimensionless spot - there is certain Focal Cloud, its size depending on used frequency of radiation.
I memorized from Internet article, that "conventional" diameter of focal cloud can be derived as a product of wavelength and certain factor or coefficient = 0.76. For 12 GHz, the wavelength = 25 mm, (1") thus FC diameter = 19 mm or 3/4".
If such a Focal Cloud radiates energy, then - due to its size - reflection from dish will be a cone, not a cylinder, as every point on dish's surface will reflect the energy according to laws of optics, and the radiation rays will spread out.
The farther the dish's reflecting spot from its Focal Cloud - the narrower the radiated cone.
Reversing the settings from radiation (illumination) to reception should also be true.
Thus John Legon was correct : wider parts of truncated dish increase its selectivity.
A sketch of drawing follows: