Alex, the following few remarks hardly deserve the status of an answer, but I think your interesting question, taken with with Greg's pertinent rider, makes it worth sharing a couple of ideas, even at the risk of downvotes from the type of churls and trolls I have already become familiar with on math stack exchange.
really I just want to relate your question to a couple of other well-known phenomena in our mental life.
firstly, I have occasionally thought that there is a curious analogy between the sufferings related to grief (whether through bereavement or various forms of traumatic separation - e.g. at adoption, or in certain cases of asymmetric relationship breakdown) and the discomfort sometimes experienced after surgical or traumatic amputation known in the literature as phantom limbs. personally I think this analogy tends to support some form of the hypothesis you propose regarding internal representations.
Greg's question is useful: how can one distinguish such internal representations from memories? of course it all depends what is meant by memory. so your question could, if necessary, be rephrased as a question about certain high-level structural aspects of memory and perception. in this context it is also worth noting that the same type of things we call self-awareness, recognition of others, and grief, have been attributed to non-human animals, particularly some species of mammals and birds.
my personal academic study of psychology took place in the early 1970's. this was an interesting time, because the intellectual bankruptcy of the still-dominant behavioristic paradigm had recently become very apparent due to its utter inadequacy to provide a basis for any but the most trivial aspects of psycholinguistics. in consequence there was a growing enthusiasm for the investigation of generative grammars and the complementary idea of species-specific biological bases for the deep structures which underlie complex aspects of perception, memory and cognition. this shift in paradigm was contemporary with a similar refocusing of interest in anthropology, which was part of a wider shift towards an approach which had become known as structuralism. those within the orbit of Anglo-Saxon empiricism were in general more resistant to this new way of thinking, but the influence was notable nevertheless.
my second point is merely a simple observation, but one whose genuineness is unproblematic in informal discourse. in personal terms, I have often noticed myself in quite casual conversation confusing the names of two former partners. to me this is evidence (admittedly only one tiny piece of a very large jigsaw) for the existence of a deep-structure corresponding to the social category of significant other. this, if it exists, is completely distinct from any kind of specific memory of individuals one has related to. of course the hard-core empiricist would still be obliged to insist this was "merely" a type of misremembering. but such a dogmatic approach ignores the insights which have long been available through the informal paradigm which underlies Freud's methodology in The Psychopathology of Everyday Life and also perhaps Jokes and the Subconscious - the general hypothesis that minor errors or distortions on the accessible surface of behaviour can offer valuable clues about the deeper mental processes which underlie the surface behaviour and play a key part in its production, but which are far less accessible to introspection. a very good example which clarifies this particular distinction is the tip-of-the-tongue phenomenon where we "know what we want to say" but just "can't find the word". often the sought-after word spontaneously arises in consciousness some time afterwards. i.e. the search, once initated, continues at the preconscious level, and this process continues to run - in the background, as it were, to use computer parlance - even though the conversation is long finished. have you ever rung a friend with some rather unimportant revelation of this type - "I've just remembered the word I couldn't recall" - only to find that the relevant context has been entirely forgotten by them?