Maddison's data should be used with some care. Besides more detailed issues about e.g. price indics applicability of a concept like GDP and, there is the fact that for a lot of countries, we lack good historical data. Such data was originally imputed and some of those values have survived. Best case scenario is that they are using proxies, but often they are just assumptions about income levels in such countries. Great care should for instance be take with all values of exactly "X00"