No, I am in fact not wrong.
The half shadow phenomenon is caused by the fact that sun is not a point source of light. The light passing through the atmosphere of Earth has a different effect, as the red light passing through and refracting and dispersing causes it to light objects in the shadow of Earth and this effect colours the Moon copper-shaded. But the geometry of shadows in space (or anywhere) is not in any way dependant on atmosphere. The shadow cast by the Moon has the exact same geometry that accounts for different kinds of solar eclipses.
The bigger the apparent diameter of the light source, the faster the shadow edges grow blurry. This image should clear it up nicely:
Or in scale:
You can see the shadow's edge getting much less defined as the distance increases, and it has fundamentally nothing to do with the atmospheric scattering.
The reason for shadows appearing so sharp and defined in space photography is because they are precisely that - at the scale that they are photographed. The length of the shadow is simply short enough that the shadow's geometry doesn't play an important role in it. Which is why I said that for general purposes, a stencil shadow system would be largely sufficient. Especially for self-shadowing of ships that are sized similar to, say, the Space Shuttle Orbiter and smaller.
But things like the Colossus are several kilometres long. At this kind of distance, some softening on the edges of the shadows would definitely happen already (depending on the apparent diameter of the light source of course). Not to mention stuff like casting shadows on other models, in which case the distances can be even larger and soft shadows (if accurately depicting the umbra/penumbra phenomenon) would definitely be beneficial to realistic lighting.
But like I said, stencil shadows would be "close enough" for most cases and definitely better than the complete lack of shadows that we have right now.
EDIT: And while it is true that far enough from the star it would appear close enough to point source of light, who says you'll always be that far from it?
You could have a situation where a red dwarf or red giant has a lot bigger diameter than our Sun at one AU distance (about half a degree). Or you could have a situation where the apparent diameter of the star is so close to point that it doesn't matter in the distances involved in FS2.
However, a stencil shadow can only deal accurately with the latter optimal scenery, while if the shadows were more accurately modeled, same system could obviously deal with both cases. It would "just" require a definition of the apparent diameter of the star's disk in mission background information. And of course a working system that correctly implements umbras, penumbras and all that jazz...