I would not discount your observations.

1 second would be twice the data allowing you to surface the raw yield data into a higher resolution surface than with 2 second intervals especially if it was harvested at high rates of speed.

Taking some good RTK elevation data and comparing that against your yield data may explain alot given the fact that you say it is very flat with undulations from tillage creating ponding.  After we vertically machine control corrected some of our flat fields we have never taken a disk into them again and that has been a good thing.

Arnie