Well, they're compositing all the pixels every frame. I don't think it's quite honest to say they're rendering 1920 x 1080 pixels per frame, as there are 960 x 1080 old pixels being 'recycled'.
I must admit that there's no strict agreement on what "render" means, so disagreement is inevitable. That said, I believe my usage is better than yours for both support and clarity. In my experience, folks typically use "render" as a general term for "what videogames and Pixar do to make images", not usually as a term for any particular step in that complicated process. The method we're discussing is certainly only a very small part of what Guerrilla's game is doing. Calling only that one step "rendering", as you would do, would mean that lighting, applying AA, and many other things done by the game are
not rendering. This seems to me both confusing and contrary to popular usage.
I'd rather use a more technical term for any specific step, and there's a very apposite one available. I'd say
Shadow Fall is
rasterizing 960x1080 pixels each frame. Now, "rasterize" is occasionally used as a general term for the whole process too. But the word is less familiar to most people, so such sloppy usage is less common. And the true meaning--calculating pixel array values from vector information--is precise enough to exclude at least some processes, making it more specific (if not quite specific enough). We can then still use "render" for the entire shebang.
So my approach is to say that
Killzone rasterizes 960x1080 pixels each frame, reprojects 960x1080 pixels each frame, and renders 1920x1080 pixels each frame (by applying lighting, AA, etc. individually to all the pixels generated by the previous two steps). That seems to me a clear and robust description. If we follow your suggestion instead, we'd have to say not only that 960x1080 pixels aren't rendered, but also that lighting is not rendered; that AA is not rendered; and so forth. I don't see what's gained by introducing all that vagueness.