When I use this script some parts of my protein appears streched.
Should I put water molecule for better refinment? Why I obtein worse
models with refine option?
Poor models are usually a consequence of a poor alignment - they will
only get worse with more refinement if your scoring function is bad,
unless you simply have insufficient sampling. Refinement does not make
your models worse otherwise, since it's using the same scoring function
as any other optimization of the models.
In comparative modeling solvation is usually implicit, in that the
solvation of the template is transfered to the target, so introduction
of water molecules probably won't help you. If you have crystal waters,
however, there may be some benefit in introducing them with appropriate
distance restraints.