I would also like an answer to this question. When I've performed PDOS calculations in VASP the projection seems in general to be quite poor so far ....

The s;pilling parameter is a measure of how well the occupied electron manifold superposes to the one spanned by the atomic wavefunctions, used to build the projected density of states. The smaller it is the better is the superposition. If the spilling parameter is large this means that your crystal wavefunctions are rather different from superposition of atomic wavefunctions.

for example the value 0.001,0.002 is that good