Residuals are not normal, potentially outliers are skewing the data.

 Non-normal residuals means the R2 is over-predicting the goodness of fit.