The deepest goal is the transfer of information from well-studied proteins
to your protein of interest.And we are going to achieve that goal by doing
sequence alignments. Thereby we assume that the sequence alignments give the
same results as we would have obtained from structure alignments. And we use
the knowledge that the location of a function in a protein is the most
conserved aspect. CLUSTAL is just software, it
cannot think, and thus cannot cope with exceptions. Humans can, at least, when
they understand the amino acids well enough to see the relation between
amino acid characteristics and alignment characteristics.