I am finding a critical difference in the pairwise alignment
computed by MODELLER and the corresponding alignment cognitive to the
CS-BLAST undertaken alignment.
That doesn't seem surprising. BLAST alignments are local and
approximate; Modeller does rigorous dynamic programming with global or
local alignments using a range of methods for constructing the residue
distance matrix.
Where CS-BLAST has got the altered sequences from, when i precisely
called the PDB for it.