During the setup of the SJT module for each question that is created the ideal 'best' and 'worst' responses are assigned according to what is expected behavior for an ideal candidate. The choices that candidates will make during the assessment will be compared to these assignments.
A benchmark can be set to have an indication of what score you would expect the ideal candidate to get. When constructing a SJT one would expect the maximum score is preferred but this score is not often achieved, even by the highest performers. So it can be more prudent to put the benchmark value lower then maximum.