There are only three possible sources of error in semiempirical methods, and if these are eliminated, then semiempirical methods would be completely accurate. These sources of error are:
(1) The reference data are inaccurate.
(2) The underlying theory is faulty.
(3) The parameter optimization is incomplete.