Dynamic Conditional correlation GARCH

[econworker]

Hi I have a question regarding to the stationary property of the series of conditional correlations obtained from a DCC-GARCH estimation. What I know is that if the sum of lambda1+lambda2<1 resulted from a multivariate DCC-GARCH model means that the conditional correlation between the series is stationary and they are mean reverting. Now suppose that I have 3 variables in my model and the results show that lambda1+lambda2<1; however when I check for stationary property of the conditional correlations obtained from the same model, the results of the unit root test shows that they are not stationary in their levels. Can any one tell me that why there is contradictory results? or maybe Im making mistake in understanding of the meaning of lambda1+lambda2<1?
I appreciate if some one help me about it.
Thanks