It is not possible to test for nutrition directly with a refractometer.
Get one for grapes and use it only to compare between the same plant materials and never to measure anything! Well, except for sugar in grapes.
The long story:
The only thing that you can do is to buy a refractometer that was designed to measure the sugar content of grapes.
Those are calibrated with the plasma of grape cells as a base value and use the refracting properties of sugar in solution added to this to approximate the sugar content.
Any additional nutrition or lack there of in the grape will interfere with the accuracy of the sugar reading.
If you measure other plants then the sugar content will interfere with the accuracy of the nutrition reading.
As you can't calibrate to the plasma of the plant cells of the plant specie you are comparing between probes right now.
The numbers will be meaningless and can only be used to compare the relative nutrition between the probes of the same plant material.
so tomato 2 has more sugar and most likely more other nutrients.
In general a higher reading means more nutrition but it will always be a total of all nutrients with a refractometer. For quantification you will need a chromatograph.
I'm planing to buy one too but I haven't settled on a make and model yet - so no buyers recommendation from me at the moment from me.
If you like this sort of shotgun comparison you could also look at the EC value. This is the conductivity and measures dominantly the content of salts in a solution. Most minerals are in solution as there salts.
Problem is, you need to juice a little cup first as EC meters need more then a drop to plunge the electrodes in to. But maybe this it totally unusable to measure mineral content of plant material because all the minerals could be bound up in enzymes, proteins, fats and sugars. On the other hand even those could carry a charge and more of them would mean more nutrition.
It sounds like you would get the most use out of a spectrophotometer: easy to load samples in solution and measure many nutrient compounds and impurities/contaminants. But this isn't strong on the cheap side or the easy side - I wouldn't recommend it as a casual interest. But if it's something you would spend a few thousand dollars plus many hours per week on, it's at least feasible. I can help arrange equipment time in Seattle for someone deeply interested in seeing this through..
So...here's something interesting. Supposed to be available March 2015.
I'm skeptical that measuring the outside surface of a fruit can give accurate information about the nutrition of the meat, but for someone that is sampling their own crop and can create better samples to scan, this might be an affordable tool.
Edit: The URL might be helpful! The product name is SCiO.