NREL 5MW control-driven instability

Dear all,
I am working on implementing state-space controllers for the NREL 5MW Turbine, based on the document (http://www.nrel.gov/docs/fy04osti/35816.pdf).
Using the 1-state model (with just GenDOF activated), I get a control gain K=-1.941. If I want to use the 3-state model (GenDOF and DrTrDOF activated) and I implement
this control gain, the control law changes the state matrix, ok? With the control gain, the state matrix becomes (A-BK). So, I should get new eigenvalues/eigenvectors
for this new matrix. In this case, I get:


My question is: the first eigenvalue has a positive real part. So, this control gain should create an instability, ok? But that’s not the case! When I do the simulations, this instability doesn’t appear.
Please help me. I am really perturbed with this apparently not logical result and I really don’t know why this happens.

Best regards,