ADF vs. KPSS

[KPSS2010]

Hi,
I have a question regarding the test results of various unit-root tests. I am using Eviews 6.0 student version.
I want to test serveral interest rate time series on stationarity or non-stationarity. I use both tests ADF and KPSS complementary(I know the difference in the H0).
ADF with AIC and mAIC automatic lag length selection says in nearly all cases time series have a unit root. KPSS test says in nearly all cases time series are stationary.

First to make sure I got the testing procedure right:
Example:
ADF with const. (AIC) = -2,5509 (crit. Value 1%=-3,51; 5%=-2,89; 10%=-2,58) ==> result time series has a unit-root even on 10% critical value)
KPPS with const. (Newey-West auto select) = 0,4335 (crit. Value 1%=0,739; 5%=0,463; 10%=0,347) ==> time series is stationary on 1% and 5% level on 10% level non-stationary (I realized with KPSS test DW-statistic = 0,038 says there is still autocorrelation, ADF recognizes autocorrelation right)

Now my questions?
1. How do I interpret the results? Which testing procedure is more reliable)?
2. While using both tests complementary, does it make sense to use different critical levels e.g. ADF = 5% level and KPSS = 10% level?
3. Could the different test results mean, that the time series has a structural break?

Thank you very much in advance for some help.
Julius