Trust-Region method does not always output the best found solution

[GrantHecht]

The trust-region method (and possibly other methods as well) appear to output the current solution upon meeting one of the convergence criteria. More often than not, this corresponds to the best-found solution, especially when convergence is stable. Occasionally, though, I've noticed that a better solution (with a lower residual infinity-norm) may have been found in previous iterations, especially if the equation is very difficult to solve and/or the max number of iterations are reached.