Human olfaction is more mnemonic because it is less structured
(Epistemic status: this just came to me on a run and should not be taken as having any particular authority. But! I asked my neuroscientist friend to explain the mnemonic character of olfaction, without giving any hint of what I thought. And she gave a very similar answer, elaborating certain parts differently. She emphasized that her answer was non-specialist speculation. But I take this experience to mean, first, this is not a particularly crazy idea. And, second, some version of this idea has probably been developed in print by professionals. I don't really know how to search neuroscience literature so instead of the version from the professionals perhaps you will enjoy this Carnap-and-Peirce-influenced version instead.)
It is often observed that smells are especially effective at conjuring vivid memories of specific occasions. Some have tried tw explain this by appeal to neuroanatomical characteristics of olfaction. I propose what I take to be a more basic explanation, in terms of the structure present in olfactory perception.
I claim that human olfactory perception is relatively poor in structure. We are able to sense a large number of different chemical signatures, but we accomplish this by means of a large number of specialized sensors, each matching a particular chemical signature. The result is that the space of human olfactory percepts does not have a rich structure of similarity relations. Scents are (often) like islands, unique and independent.
Note also that our olfactory perception is not spatially structured, unlike other animals' olfaction and unlike our own vision. Our olfactory percepts are single-pointed -- we smell a smell (or smells), not a spatial array of smells. For comparison, imagine that our vision was limited to a single "pixel," that we "saw" red or green or blue or white single-pointedly, instead of at many individual points in a larger array. Imagine further that we didn't compute from our vision a spatial map of light emitters and reflectors. Then our visual experience would have as little spatial structure as our olfactory does. Spatial structure allows us to compute similarities between percepts that differ in all non-spatial respects. For example, we can recognize that a head made of flesh and a head made of vegetables are alike in spatial distribution, even though they are very different at every point. Because they lack spatial structure, such similarities are much less available as between our olfactory percepts.
The space of our visual percepts is characterized by a very rich similarity structure. Any given percept can be similar to many others because they can be alike in local color or texture or brightness, or in any number of global attributes, like shape or brightness or overall color. In contrast, the space of our olfactory percepts is much less connected by similarity structure. Many possible olfactory percepts are only similar to relatively few others. Some olfactory percepts may only be similar to other events of smelling the same rare thing.
The contrast in the mnemonic significance of olfaction and vision can then be explained with a simple auxiliary assumption. Suppose that, when a percept is perceived, memories of episodes containing similar percepts are activated, in proportion to degree of similarity. For a typical visual percept, similarity is distributed widely so no particular memory is strongly activated. For an olfactory percept, if there are similar percepts in memory, they will often uniquely pick out an episode in memory, resulting in that particular memory being strongly activated.