SVAR: I am puzzled

[QSnakecharmer]

Ok, I really need help friends...

(1) When I ask Eviews to estimate a six variables' SVAR(p) of order p=1 with a long-run triangular matrix (with zeros above) "a la" Blanchard-Quah, I receive this ERROR MESSAGE from Eviews:

"Hessian matrix is near singular at final iteration parameter values"

(2) But, when I ask Eviews to estimate the same six variables SVAR(p=1) with the same long-run triangular matrix, but with normalized variables in this case, there's NO error message.

If I want to see the IRFs of the SVAR with the variables expressed in his originals units of measurement (the one in 1) and NOT the normalized (2)... what can i do?

Little help, please.

P.S.: By the way, I love Eviews. Its better than playstation :D