Time slice job will not run when it is already running ; so if the job has more than "1 minutes worth" of data to slice then you would see this sort of behaviour.
i.e. if the job takes 5 minutes to process the data, you would see a gap of 4 minutes between executions (last updated date on the slice table).
Also, slight discrepancies in time stamps can be explained where there are discrepancies in the "current time" between the various servers in your system - i.e. if the database server's time stamp is not exactly synchronised with the application server's time stamp - it is hard to tell which time stamp is displayed to the user (usually the database time stamp I think) but the application time stamp would be used for the schedule of when jobs execute.