I'm pretty sure that I saw that as well, but there is a further level of abstraction between a scale model and a map. A scale model of a room is considerably more like the room itself than a plan of the room is.
One recent similar scale model based experiment with chimpanzees is documented in
Chimpanzees (Pan troglodytes) recognize spatial and object correspondences between a scale model and its referent.Kuhlmeier VA, Boysen ST.
Psychol Sci. 2002 Jan;13(1):60-3.
n the present study, the contributions of spatial and object features to chimpanzees' comprehension of scale models were examined. Seven chimpanzees that previously demonstrated the ability to use a scale model as an information source for the location of a hidden item were tested under conditions manipulating the feature correspondence and spatial-relational correspondence between objects in the model and an outdoor enclosure. In Experiment 1, subjects solved the task under two conditions in which one object cue (color or shape) was unavailable, but positional cues remained. Additionally, performance was above chance under a third condition in which both types of object cues, but not position cues, were available. In Experiment 2, 2 subjects solved the task under a condition in which shape and color object cues were simultaneously unavailable. The results suggest that, much like young children, chimpanzees are sensitive to both object and spatial-relational correspondences between a model and its referent.
I would agree that if the mouse behaved as you described it would suggest that it was effectively interpreting the map. Perhaps a mouse in a maze situation would makes it harder than it needs to be though, mazes are pretty tricky to negotiate anyway.
TTFN,
WK