When running a configuration with a warm-up period workgen accumulates the operations done during the warm-up period. Therefore the statistics reported in workload.stat for the first report are wild outliers.

The program should either not accumulate the operations during the warm-up, or zero them out at the end of the warm-up period.

Also the `(120 total secs)` is in error because it really should start at 0 (or 1) when it reports its first set of statistics.

Here's an illustrative example:

449362 reads (9 not found), 243556 inserts, 0 updates, 0 truncates, 0 removes, 0 checkpoints in 1 secs (120 total secs)
1396 reads, 1720 inserts, 0 updates, 0 truncates, 0 removes, 0 checkpoints in 1 secs (121 total secs)
3021 reads, 2054 inserts, 0 updates, 0 truncates, 0 removes, 0 checkpoints in 1 secs (122 total secs)
...