Test plots are put on the ground that are usually somewhat good.  You wouldn't put a test plot in a water hole or the top of a sand dune.  All those areas do better and then you have higher national yield. 

 Too few of data points.

3 plots in Kansas is way to few for a state that large and so variable, without knowing where in the state those plots are it is impossible to know if the lower yield between those years was because it was too wet or too dry.