That's my whole point. In smaller tests, this is true. In a 45 minute test, the elapsed time is greater than the sum of the duration. However, on longer tests (4.5 hours) the elapsed time is correct, but the sum of the duration far exceeds the elapsed time.

As Stefan_G suggests, if the problem is that transaction 1 starts at 00:03:00.000 and ends at 00:03:01.000, and transaction 2 starts and ends at the exact same time, the total duration is 1-second, but the sum of the duration column would be 2-seconds. If it were this simple, I could figure out a way to determine the true duration time. However, it's unlikely that they would start and end at the exact same time.