Correction author: David Pearlman
Correction date: 8/91

Problem: To work correctly, the internal restraint routines in SANDER
  expect that the inflection points of the restraint well, r1->r4 and
  r1a->r4a, be monotonically increasing. If they are not, erroneous
  energies could be reported.

Symptoms: Negative energies could be reported/calculated for restraints
  where r1->r4 are not monotonically increasing. 

Other notes: It is conceivable that one might inadvertently specify a
  non-monotonic series for a torsional restraint. E.g. one might assume
  it was reasonable to specify r1 = 330., r2 = 345., r3 = 15.0, and r4 = 30.0
  to set up a restraint centered on 0.0 degrees. But this would lead to
  erroneous energies. The correct way to specify this restraint would be
  to use the series: r1 = -30., r2 = -15., r3 = 15., r4 = 30.

Workaround: In the 4.0 version of SANDER4, the user must ensure for 
  him/herself that each restraint is specified by a monotonic series.

Fix: A check has been added to routine NMRRED which forces an error and
  program termination if a restraint is not specified as a monotonically
  increasing series.

Routines affected: nmrred.f.