http://amasci.com/amateur/holo3.html
Having studied with Steve Benton at the Media Lab and built some of the highest fidelity white light 3D displays of the time, I would agree with this article: scratch holograms behave very similar to rainbow ("Benton") holograms. They use a combination of diffraction, diffusion, and parallax to produce an HPO (horizontal parallax only) 3D image.
Yes, it isn't made by interference. No, it isn't full parallax or does it form an image purely by diffraction. But a scratch hologram is the superposition of the equivalent of many zone plates, and the basic mechanics are the same as in a hologram (well, a holographic stereogram). It definitely isn't "just" parallax.
All that said, I agree that what's happening is very different from a computed hologram that models interference mathematically and records the detailed pattern on a recording material. Computing and writing such a pattern requires very high resolution to get the high spatial frequencies needed to get large diffraction angles. That makes it very tedious to make any image, and even worse one made of lots of points.
A scratch hologram (or a holographic stereogram) gets around that problem by essentially modulating details in the spatial domain (the image) onto a high frequency diffractive or diffusing carrier signal (the scratch, or a holographic diffraction pattern in a holographic stereogram).