The Journal, Timelog and Timedot readers' detectors now check each line in the sample data, not just the first one. I think the sample data is only about 30 chars right now, but even so this fixed a format detection issue I was seeing.