Hello
Vacation is over so its time to answer some questions
You probably used a Hessian update method within statpt. This results in only an approximate
Hessian matrix which can well have several negative eigenvalues, which most probably
will not be the same as calculated with analytical (aoforce) Hessian.
Did you calculate Hessian with aoforce prior to running TS optimization? If nothing helps
you can try to run statpt with analytical Hessian calculated every 3-5 steps. This will be
expensive but should lead finally to TS.
The step size in statpt is adjusted automatically. You can only specify initial and maximum
step size (using define menu). Try to decrease the maximum step length, it may also help.