Hi,
I did give a quick explanation and gave some reference on the subject in the following forum thread Question about mono-EL7 and shared-EL7 partitions usage .
But in short the estimated wait time is very pessimistic it is based on the maximum time a job is announced to take. As the default is given by the maximum time limit of the partition which is 4 days and more in some partition the predicted time is also use. To be more precise a job with a maximum run time of 4 days will be sheduled to take the whole 4 days and other jobs will be sheduled to run after it has finish. However if the job only take 10 hours the other job can start sooner.
In short inaccurate wait time is due to overestimated time limit. Tighter your time limits are, more accurate is the sheduling and faster your job will start. Of course by design your time limit should always be bigger of what you need to avoid having your job kill for time limit.