Are both controllers connected to both fabrics?
From your description it doesn't sounds like it matters if you go over fabric A vs B, it's which controller creates the LUN that seems to make the difference here.
The usual stuff to check is the active zoneset and make sure that both initiators (one for A and one for B) are allowed to hit the respective ports for the A and B controllers on the Netapp.
If the blades are supposed to boot from these luns then you can SSH to the UCSM address and after a reboot run the following to get a report from the option ROMs that should tell you more about what's wrong:
connect adapter x/y/z (chassis / blade / adapter)
connect
attach-fls
lunlist
lunmap
What is going on on the controller side?
What do you get if you run "igroup show -v"?