Dear Yulin,
Your linearization process sounds reasonable.
I believe there is quite a bit of coupling between the tower-bending modes and the rotor blade fexibility for this model. This coupling will have a small impact on the 1st tower-bending frequencies and a large impact on the 2nd tower-bending natural frequenices. I suspect you’ll be able to match the results to NREL/SR-500-45891 much better if you enable the blade-bending DOFs in FAST. More information is available in the forum topic found here: Tower Eigenfrequencies of NREL 5MW Turbine.
Best regards,