I use Modeller for about two years and I can rarely
set up a .PIR file and run the program with no errors
for the first time (and have seen many times questions
on alignment errors here). This is not really
Modeller's manual fault - I think that what causes
most of those errors is that the sequence listed on a
PDB entry does not always correspond to what is really
on the PDB file. Some entries include, for example,
His-tags which are never visible and also residues
which could not be seen on electron density maps and
therefore were not modeled. Setting up the .PIR file
sometimes means to search for missing or modified
residues (e.g. Se-Met or modified Cysteines) on a PDB
and then fix the alignment file, something that can be
really boring in some cases. I have developed a "semi
automatic" approach to do that, but perhaps it would
be interesting to include on newer versions of
modeller a routine to parse the PDB, compare it with
the sequence on the alignment file (which may include
these missing residues) and fix the file.