The drift correction is applied in the concentration calculation (there is an explanation of this equation in the basics webinar). This calculation uses the spline value of the reference materials.
If there is still drift in your samples, it means that the spline describing the primary reference material is not accurately modelling the drift. This can happen especially if you've chosen something like a linear fit model, or some other simpler model.
If your primary RM does not have any residual drift, but your secondary RMs do, this suggests that there is some other effect decoupling your secondary RMs from your Primary RM. This latter case is pretty rare though.
If that doesn't help, please let us know.