The wait times specified for reservations are estimates only. These estimates are based on average session lengths to try to give the most likely wait possible. Essentially, SignUp calculates the average session time based on past sessions, and uses this value to guess at which current session will end next. The predicted end time for this session is then used to inform users when the next computer will be free. However, if the session that is predicted to end first actually goes on for longer than average, the calculation will be performed again - this often results in the wait time increasing. On there other hand, if another session ends early, the wait time immediately drops.
Pharos considered options such as simply displaying the wait time based on the assumption that patrons will use their full sessions; however this is not always the case, as sessions can end early. If patrons were only presented with a specific session start time, and someone were to end a session early, then their session could start considerably earlier, but it's likely that the user would have seen the later start time and walked off, missing their session start.
Some control on how the estimation engine determines these 'estimated times' can be obtained by modifying the Signup registry key (see below). By increasing or decreasing this value you can affect the accuracy of the estimated wait times. Times will still vary due to session end time differences, but decreasing this amount should reduce the amount of that variance.
The key is [HKEY_LOCAL_MACHINE\SOFTWARE\Pharos\SignUp Server\Minimum Sample Duration Percentage]
NOTE: Additional information regarding this key can be found in Pharos Help by searching for "SignUp Registry Entries."