My first inclination would be to base maintenance cost off cost of rebuilding (optionally overridable) -- in effect, you end up with a much lower percent of total value where you're basically paying for an expensive lot, and a much higher percent of total value in depressed markets (e.g. where you can get 100-year-old 7-bedroom for ~= national median home value).

Just guessing here, would it make more sense to base yearly maintenance on $/sq ft instead of overall property value? Would better take in to account expensive but modest (lower maintenance cost) versus huge home (higher maintenance cost).

$/sqft glosses over local labor costs, level of finish, landscaping, etc., which is why you see people suggesting 1-2% of value per year, though you're right that it would be an improvement to separate out the value of labor/materials from the value of a the bare lot.

Just draw the user's attention to this and any derived/estimated values. It's impossible to model this correctly without very specific regional data.