Long run restrictions in SVAR

[03604808]

Dear all
I'm using EViews Version 5.1 and need to estimate a SVAR with long run restrictions à la Blanchard & Quah.
To do so I first estimate a VAR, then I estimate a structural Factorization with long run restrictions (I specified a pattern matrix). Finally I look at the IRF using the structural decomposition method.

This doesn't yield the expected results, furthermore I checked the variance of the structural shocks (u) and this did NOT yield the identity matrix. But this should held by construction (Model: Ae = Bu where E[uu']=I)!

Did anybody already encounter that problem? Maybe there's an error in my specification?

Thank you for your help!

...