I have to say that I've never used any of the QM/MM methods myself, so I cannot comment on the QM setup.
However, I've noticed a few things in the output log that may be relevant.
First, the CONS HARM command may need to follow the reading of the coordinates from ts254.crd, esp. if the restrained coordinates are significantly different from those in the initial coordinates read earlier in the script.
Second, at dynamics startup the message
**** Warning **** The following extraneous characters
were found while command processing in DCNTRL
TREF 300
indicates a failed attempt to establish a thermostat. That is the T setting for the weak coupling (Berendsen) thermostat, which is known to be flawed; it gives a T distribution which is significantly narrower than a proper Boltzmann distribution, and its usage is deprecated. The TCONS and TCOUPLE keywords are also required for that thermostat.
Third, I note there are no boundary conditions of any kind. While this may not a huge issue with the QM part, there are known artifacts from the particle:vacuum interface present for a system with no periodic boundary conditions. The more rigorous Hoover thermostat requires the use of periodic boundary conditions via CRYSTAL commands.
