Ian,
Configuring the hardware for continuous will not make a difference. All the configuration will be done before the task is started so this will not create anymore offset. I saw similar results when I set up a "Tee". The phase difference of the samples themselves were only separated by a few microseconds.
I did, however, see quite a difference in the t0's. This is caused by the execution of the VI's starting the task being software timed. You can control this somewhat in LabVIEW using a sequence structure, but it still could vary depending on the resources and usage of your computer. This could cause around 1ms or even more of a difference.
Regards,
Chris Delvizis
National Instruments