I think it's more likely to be a design decision. If you try to model a large real world area, some people will immediately notice how it is unlike reality; so the constraints of computational power, design skill, and developer ability are multiplied. The differences can be truly jarring.

My guess is that there is more upside to creating unique areas. This affords more creative freedom and the luxury of changing the simulation in accordance with constraints and resources.