Problems with Unit Root Tests

[Tisoon]

Hello,

I want to estimate a ECM or VECM model and I must to start to control the order of integration of (3) variables. There are the sells of diesel (lgz), the GDP (lpib) and the price of diesel (lpx).
But I've problems and understanding.

For the variable LGZ :
http://img515.imageshack.us/img515/6340/lgzn.jpg
I don't understand why I find good results for ADF test in level and not in first difference. In level, the variable grows continually and doesn't seem stationary, in consequence.
For Philips-Perron (p=5) test, result is good in first difference only (with constant, for 5% it's 2.46 < 2.52, and so the perfect result is with no constant and no trend).
Can you explain me why ADF test has this result? Is it false, yes?

For the variable LPIB :
http://img13.imageshack.us/img13/811/lpib.jpg
ADF and PP (p=5) tests have the same result. For me, it's stationary in second difference. But my professor says that it is false and it's I(1) for LPIB.
"The LPIB variable is I(1), it's a general result that we find in the literature and your study."
But I don't understand, completely, why? And how? The graph in second difference seems very good.

I've no problem with LPX.

Can you help me, please? Thank you very much for your assistance.

If need, my Eviews file :