In some agencies, it goes without saying that transit maps should be geographically accurate. Many agencies follow San Francisco Muni in superimposing transit lines on a detailed map of the city:
But research out of MIT suggests that we really need to see network structure, and that requires a degree of abstraction:
By putting alternate versions of the New York and Boston subway maps through the computer model, the researchers showed that abstract versions of the maps (as opposed to geographically accurate versions) were more likely to be easily understood in a single, passing glance.
Geographical accuracy obscures network structure. Purely geographic maps show where service is but not how it works.
This is why a number of best practice agencies publish both kinds of maps, sometimes even presenting them side by side. The geographic map helps you locate yourself and points of interest in the city, but you need the structure map to understand how the system works.
All this is even more urgently true for bus network maps, where complexity can be crushing to the user. When we streamline maps to highlight key distinctions of usefulness such as frequency, we often have to compromise on geographic detail. Obviously the best maps fuse elements of the two, but you can always find the tradeoff in action. The new Washington DC transit maps, for example, highlight frequency (and show all operators' services together) but there's a limit to the number of points of interst you can highlight when keeping the structure clear: