Refractometer use.

First up it should be said that before anything else, salinity should not be checked unless all waters including RO water for are kept at approximately 25c to get a correct reading.

These words are kindly given by Randy Holms-Farley.
It is always correct to calibrate reef refractometers using a proper 35 ppt (1.0264) standard to make measurements of seawater.

It can be inherently inaccurate to calibrate at 0 ppt (RO water) because these are often brine refractometers and not actually seawater refractometers. Since that is often the case, if you calibrate at 0 ppt, you will read 35 ppt incorrectly even if the refractometers is perfectly made and working perfectly.

I discuss that here:

Refractometers and Salinity Measurement
http://reefkeeping.com/issues/2006-12/rhf/index.php

From it:

Imperfect Refractometer Use: Scale Misunderstanding and Salt Refractometers
Refractometers can lead to incorrect readings in additional ways and, again, these issues abound for reef aquarists. One is that many refractometers are intended to measure sodium chloride solutions, not seawater. These are often called salt or brine refractometers. Despite the scale reading in ppt (‰) or specific gravity, they are not intended to be used for seawater. Unfortunately, many refractometers used by aquarists fall into this category. In fact, very few refractometers used by hobbyists are true seawater refractometers.

Fortunately for aquarists, the differences between a salt refractometer and a seawater refractometer are not too large. A 35 ppt sodium chloride solution (3.5 weight percent sodium chloride in water) has the same refractive index as a 33.3 ppt seawater solution, so the error in using a perfectly calibrated salt refractometer is about 1.7 ppt, or 5% of the total salinity. This error is significant, in my opinion, but not usually enough to cause a reef aquarium to fail, assuming the aquarist has targeted an appropriate salinity in the first place. Figure 23 shows the relationship between a perfectly calibrated and accurate salt refractometer and a perfectly calibrated and accurate seawater refractometer when the units are reported in salinity. This figure shows the measured salinity reading for seawater being about 1.7 ppt higher than it really is.

It turns out that this is a slope miscalibration in the sense that a perfectly made sodium chloride refractometer necessarily has a different relationship between refractive index and salinity than does seawater. This type of problem with a refractometer IS NOT at all corrected by calibrating it with pure freshwater. If you have this type of refractometer, and it was perfectly made and calibrated in freshwater, it will ALWAYS read seawater to be higher in salinity than it actually is (misreporting an actual 33.3 ppt to be 35 ppt).

Even more confusing, but perhaps a bit less of a problem in terms of the error's magnitude, salt refractometers sometimes read in specific gravity. But that value is specific gravity of a sodium chloride solution with the measured refractive index, not seawater with that refractive index. A sodium chloride solution with the same refractive index as 35 ppt seawater (which turns out to be 36.5 ppt sodium chloride) has a specific gravity matching 34.3 ppt seawater. So this type of refractometer, when perfectly calibrated, will read the specific gravity of 35 ppt seawater to be a bit low, at 1.0261 instead of about 1.0264. That error (reading 0.0003 or so too low) is, however, probably less than most reef aquarists are concerned with. Figure 24 shows the relationship between a perfectly calibrated and accurate salt refractometer and a perfectly calibrated and accurate seawater refractometer when the units are reported in specific gravity. This figure shows the measured salinity reading for seawater being about 0.0003 lower than it really is.

Regardless of a salt refractometer's scale reading (ppt or specific gravity), aquarists can get around this problem by calibrating this type of refractometer in a seawater standard, Because that type of calibration also gets around important manufacturing errors (slope calibration defects due to the scale being the wrong dimensions), it solves both problems at once.