Cluster expansion with vacancies

I tried to include vacancies in some sites, and it gives str.out as below:
3.600000 0.000000 0.000000
0.000000 3.600000 0.000000
0.000000 0.000000 3.600000
-0.500000 -0.000000 2.500000
0.500000 -0.000000 2.500000
0.000000 0.500000 -2.500000
0.000000 0.500000 2.500000 Vac
0.000000 0.500000 1.500000 Vac
-0.000000 0.500000 0.500000 Vac
0.000000 0.500000 -0.500000 Vac
0.000000 0.500000 -1.500000 Al

By using vasp glue, it gives the POSCAR.static as below, which will gives error when running.
title
1.00000000000000
-1.8629751907496794 0.0000000000000000 8.9986309544583545
1.8629751907496794 0.0000000000000000 8.9986309544583545
0.0000000000000000 1.8629751907496794 -8.9986309544583545
Al
1
Direct
0.2000000000000028 0.2000000000000028 0.0000000000000000

0.00000000E+00 0.00000000E+00 0.00000000E+00

Does anyone know how to handle this?

I would be useful to know exactly what is the error message.
I am guessing that VASP is complaining because the structure has a lot of empty space, so it shrinks a lot during optimization and the planewaves just become poorly optimized.

When you have vacancies, it is important to limit the fraction of vacancies in the system (otherwise you get nonphysical structures with a lot of empty space). This is done by specifying a crange.in file (for mmaps.in ) Please see mmaps -h for the syntax.

Can I define the fraction of vacancies for each sub-lattice?

Sorry, this is not implemented - only the overall concentration can be controlled.
Now, you can always chose to not compute those structures that have too many vacancies on a given sublattice. mmaps won’t complain. Just remove the "wait" file and do "touch error" within the structure’s directory and mmaps will go on.