Isn't the whole question of whether it's a reflection as from a conducting
plate, or radiation from the radials based on misunderstanding what a
reflection is?

The reflection from a conducting plane is the sum effect of the radiation
from the currents induced in the plane by the incident field, say from a
quarter wave vertical. The reflection forms an image 'behind the mirror' of
the source of the incident field, for example a quarter wave vertical 'under
the plane'. If these currents are radial then the plane is well approximated
by a large number of long radial wires. The high resistance at right angles
to the radials doesn't matter because no current is flowing in that
direction. The plane is not so well approximated by a small number, say 3 or
4, radials, especially if they are short, say a quarter wavelength.

However the approximation may be good enough to allow us to regard the short
radials as reflecting the incident field from the quarter wave vertical
especially if we just need a rough picture of the situation and a rough
estimate of the input impedance.We may need to keep in mind that the
reflection is due to radiation from the currents in the radials or we may be
able to forget it. If we want the antenna pattern in detail and the input
impedance accurately we have to do the maths properly. That's not simple
even for a half-wave dipole!

Brian LA0DG