This is a generic answer, but the key really is 'scope'. If you only have 2 sites, then presumably those sites are specific in terms of (something like) demographic, location, discipline, behaviour, etc. If you refine the scope such that you're looking at this specific demographic (etc.), then it generally makes it easier to explain the context of the findings and the limitations in terms of generalisability.
There's a good PhD comics on the 'research' method (where you define the question, then ascertain a methodology, etc.); and the 'actual' method, where you end up reverse engineering the research question to fit pragmatic constraints. Not an ideal, but often a reality.