It looks like CP with LDA+U gets confused when there are atomic wavefunctions

without U with the same l of atomic wavefunctions with a Hubbard U (example:
TiO2 with U on Ti(3d), ok; FeCuO_2 with U on Fe(3d), not ok). Setting Hubbard_l
to -1 for non-Hubbard atomic wavefunctions should do the job. No warranty.

CPV/src/ldaU.f90

@@ -84,6 +84,8 @@
             Hubbard_l(is) = set_hubbard_l( upf(is)%psd )
             Hubbard_lmax = max(Hubbard_lmax,Hubbard_l(is))
             write (6,*) ' HUBBARD L FOR TYPE ',atm(is),' IS ', Hubbard_l(is)
+         else
+            Hubbard_l(is) = -1
          end if
       end do
       write (6,*) ' MAXIMUM HUBBARD L IS ', Hubbard_lmax