pdb-Files made by GROMACS with more than 9999 residues
(e.g. Protein in Waterbox) have selection-problems.
After residuenumber 9999 it begins again with number 1.
So, you have for example different water-molecules with
the residuenumber 1234.
Now the problem:
select bla, all within 3 of r. GTP
does not only pick the atoms around the GTP. It selects
all other water-atoms with the same residuenumber (even
if there is a distance ov more than 20 to the GTP).
You can test this by using my attched pdb-file.