Hi chenhongjun ,
We received your email, and we've had a good look at the data you sent. In general, the data look good. The secondary reference materials all come out within uncertainty of their accepted values. The uncertainty is perhaps a little large, but this could potentially be reduced by running more analyses of the primary reference material. Given that the downhole fractionation correction is the largest correction made to the data, the downhole fit is very important, and the more analyses of your primary reference material you can use to determine the downhole model, the better this correction will be.
Regarding which model to use for the downhole fit (exponential etc), I processed your data using the exponential fit and it worked pretty well (208/232 could perhaps have been linear, but not a problem there). The best way to tell if a model is working well is looking at the residuals (the bottom graph). If this line is centered on zero, without a strong trend in the data (i.e. not consistently above or below zero), then it suggests your model is OK.
I hope that helps? Please let us know if you have any other questions (others might have the same questions).
Bence