High dielectric samples (e.g. salty samples) dissipate the RF energy very efficiently. This is the reason for the longer pulse widths observed by you.
Water suppression in salty samples by presaturation causes extensive heating of the sample, due to dissipation of the RF energy within the sample.
sample. The resultant thermal gradients are an additional source of poor water suppression as the chemical shift of water is sensitive to temperature.
However, you are not using presaturation. Based on your observations, I would guess that the problem lies in the non-linearity of the RF pulses due to power dissipation within the sample, IF the excitation sculpting scheme involved proton RF pulses of different powers (soft pulses) in the watergate sequence. If only hard pulses were used for excitation sculpting then this is not the primary source of the your problem.
The second problem is the loss of magnetic homogeneity due to introduction of a sample with high dielectric medium. Though the line-width at half height may not change significantly, the peak shape may be distorted. Even minor distortions in the peak shape can lead to very poor water suppression, even with a watergate based solvent suppression scheme. Therefore, the shimset used for this sample has to be optimized extensively and will be different from that of other samples in routine use.
