As is reflected in the answers posted by others the normal expectation is that the CB is set based on some logical basis, usually a coordination study performed while considering expected loads, transient loads, cable and equipment ratings, etc. And the AF study is performed using those settings. Possible iterating the process if Ei results are undesirable and can be improved by slightly different OCPD settings or selections.
In modern adjustable CBs the settings should be lockable and require a password or a deliberate action to bypass the protection. If,in the subject facility, there is a real risk that personnel are changing settings for whatever reason, not recording those changes and the people performing tasks or exposed to the hazard are not aware, then it may be prudent to take worst case into consideration.
It comes down to risk analysis and risk management. The common experience is that the desired settings are the settings and it flows from there, but if there is a risk or probability that that is not the case then "you" need to decide if the residual risk is acceptable and if not implement an action or control that reduces the residual risk to an acceptable, at a minimum. Using worst case EI numbers may be one way, another may be to implement administrative procedures that verify settings before a worker executes task or approaches the hazard, etc. What ever you do needs to fit within the electrical safety policies and procedures of the facility, be reasonable within the context of the skill sets of the workers potentially exposed, fit within the context of how equipment is maintained and operated, etc.
Hope this helps... I think the answer that I provided, in short is... it depends!
