See my response to your post in the GPC section for more background. In GPC, elution order is base on size, with small molecules eluting last. The diluent solvent molecules are small, so any peak that they form will be at the end of the chromatogram. In ion exchange, the diluent solvent molecules are usually un- or weakly retained, so any system peak formed will occur at the beginning of the chromatogram -- that's what the "water dip" is. It's usually negative because the diluent at that point no longer contains many ions, hence has lower conductivity.
To a first approximation, affinity for the resin is related to the charge density of the *hydrated* ion; it increases as you go down the periodic table. Visualize it as fluoride having a lot of water stuck to it, chloride less, and so on.