The study shows that Johansen's tests for cointegration can be less accurate in certain situations, like when there are extra unrelated factors affecting the data, or when there are multiple cointegration relationships at play. This means that using these tests on systems with both real and nominal data can lead to incorrect results. To get more reliable results, it's better to test for cointegration in smaller systems where economic theory suggests it should exist. The study provides examples where not following this advice led to cointegration not being detected in various economic relationships.