With the concept of mapping via abstract higher-order relations like "cause" (and here we can think about the whole range of functions in engineering, such as "transform", "import", etc.), we can begin to see why far analogies might be the most useful and produce the most interesting creative leaps. I think part of the story is that, all things being equal, the nearer the analogy, the greater the overlap in knowledge between the target and source, and thus the lesser the opportunity for interesting alterations to the target representation.
There is some experimental evidence that more directly measure the creativity impacts of mismatches in attributes. These studies generally find that far analogies that mismatch core attributes of a problem do not lead to creativity gains.