I think also that it looked ok. I assume that you want to display the sum against any higher dimensionality as the employee and workdate. Just reduce the dataset per selections to two of them and remove then step by step the more granular dimensions - normally it should work and returning the appropriate sums. Helpful might be also to keep the lower granularities and to add some TOTAL statements to ignore certain parts of the object-dimensionality.
If not you may try to extend the aggr() with a NODISTINCT and to integrate some kind of error/exception-handling, for example to wrap the single and/or the combination of the min/max within an alt() or a rangesum() because each non-numeric parameter of a calculation with normal math. operators will return NULL. Useful might be also the attempt not to calculate the target at once else to return the underlying values which may be done per:
concat(aggr(max(S_date_and_time) & ' - ' & min(S_date_and_time),S_employee_id, WorkDate), ' + ')
By everything be not too fast to discard an unexpected result because it may return valuable hints to the underlying data and why the intended approach may not be working.
