The principle of isomorphism states that the perceptual mechanisn is in every way similar to observed properties of the subjective percept. An isomorphic model therefore cannot fail to describe the phenomenon that it models. If one perceives an edge completing amodally behind an occluding surface, then the isomorphic model must also complete the edge behind the occluding surface. As "obvious" as this might seem, most other models are based on the assumption that the internal representation of the perceived occlusion is something less than a fully spatial replica of what is observed perceptually, usually being defined by computation in a two-dimensional plane. The central point of this modeling approach is exactly that an isomorphic model must practically amount to a description of amodal completion.

As to "how such a propagation takes place", this model states explicitly that it occurs as a spatial diffusion of a surface signal in a volumetric spatial representation, which encodes both the modal foreground surface as well as the amodal occluded edge. That aspect of the model is both explicitly stated, and (to my knowledge) unique in the literature of computational models of perception. This is not just a description of the phenomenon, but a statement that the perceptual mechanism itself is similar to the phenomenon, i.e. that a description of the phenomenon is also a description of the perceptual mechanism while perceiving the phenomenon, as suggested by the principle of isomorphism.