When calibrating a refractometer with 35 ppt solution, should the refractometer then ideally be at exactly the 0 line with RODI water or below the 0 line?
Many refractometers used in the hobby are brine refractometers repurposed from other industries to sell to reef hobbyists. I suspect many resellers do not even understand what they are selling and why it is not ideal.
A brine refractometer is made to measure salinity of sodium chloride, which is a lot, but not all of what seawater is.
A true seawater refractometer is designed to measure the salinity of seawater.
A perfectly made and perfectly calibrated brine refractometer cannot read both zero ppt and 35 ppt correctly. One must be off, as you have found. If you calibrated with an accurate standard at 35 ppt, then it will read 35 ppt correctly, and will even do so if less than perfectly made.
A perfectly made and perfectly calibrated true seawater refractometer can read both zero ppt and 35 ppt correctly. If you calibrated with an accurate standard at 35 ppt, then it will read 35 ppt correctly, and will even do so if less than perfectly made.
Thus, there are several scenarios where calibration with 0 ppt solutions can lead to errors, and no scenarios where calibration with an accurate 35 ppt (or similar salinity) standard will lead to errors.
That said, commercial salinity standards seem to have a poor track record of accuracy for reasons that I cannot understand. They are cheap and easy to make. Why do they go wrong?
For this reason, I use and recommend a diy using a good scale for both the sodium chloride and the water. Using a salt mix to make a standard is also not a good plan since it can have variable amounts of water in the solids.