Setting Constraint Tolerance while using pattern search for a constrained minimization problem

I am solving a constrained minimization problem using pattern search.

While setting max constraint violation, I was wondering what its ideal value should be. When I keep it less, I get values of the objective function less than the values I get using local optimization solver such as fmincon(which is desired, considering pattern search as a global optimization algorithm). But when it is kept to the default 10^-6, I get values which are more.

Could anyone suggest whether I can go with setting lesser values or maybe suggest what the ideal value should be for max constraint violation.

[Mario Malic]

Having no information about the actual problem, it's hard for us to suggest anything. Your problem is constrained, will it make sense to violate the bounds you have set, do you expect that your objective could have a minimum around the bounds? Try surrogate solver if you have a bit more variables if it's suitable for your problem.