Semicore band ranges too large

Explanation about this warning by Peter Blaha on the wien2k mailing list (Nov 5, 2013):

This new test (v 13.1) checks for ghostbands, which otherwise would go through without notice. I consider it a VERY important check and in almost all cases it is a severe problem.

It has the same origin as the STOP due to "QTL-B too large", namely unphysical ghostbands.

Ghostbands occure for two reasons:
a) the energy parameters are not set properly. This means most of the time that the wien2k-defaults for your system (mainly due to the required RMT values or due to a particular charge transfer) are not good. In most cases it comes because for ONE atoms there are 2 expansion energies for the same "l", and these energies are too close.
b) The sphere sizes are too different. This happens in particular when a cation (like your Na or Ca) has too large spheres compared to H, O, B. Often such ghostbands appear for (partly) converged calculations. In that case, you'll probably have to reduce the larger spheres.