Rapidminer go: Prediction derived from model simulator different than when run as model

cramsden

I am using Rapidminer GO, and have just come across something very confusing.   I have a model built using gradient boosted trees.  I have a new data set that I am using that model to get a prediction on.  I inputted the new data into the 'model simulator' as well as just using the model on the new data set.  I get completely different prediction result when using the simulator vs apply the same model to the same data set.

How can this be possible?  I checked multiple times to make sure the data being inputted into the model simulator was the exact same as the data in the csv file but the prediction results are different.