I've been looking into the role that analogies play in scientific discourse-- trying to distinguish the interactive, dialogic role of analogies from the more cognitive, logical problem-solving role that characterizes much of the research on analogical reasoning.
Clark and Shaefer's work on grounding in discourse is a possible direction for understanding analogies:
When people take part in a conversation, they bring with them a certain amount of baggage— prior beliefs, assumptions and other information. Part of that baggage is their common ground, which Stalnaker (1978) described this way: “roughly speaking, the presuppositions whose truth he takes for granted as part of the background of the conversation… Presuppositions are what is taken by the speaker to be the common ground of the participants in the conversation, what is treated as their common knowledge or mutual knowledge.” (p. 320, original emphases)… in actual conversations, the presuppositions vary from one participant to the next, though usually not too drastically. (p 260 of C & S)
Analogies can perhaps be understood as a move to establish common ground-- choosing a domain in which the participants can agree on all the presuppositions regarding the structure of the domain and the kinds of ontologies involved. Of course, the participants might disagree on the appropriateness of the analogy (might be a kind of structure-mapping move as this happens). Most of the analogies in my corpus (of spontaneously generated analogies in science discourse) are not used to map but to ground.-- Once the analogy is agreed on as appropriate, it has served its purpose and is no longer used.