Here's what I know based on my experience. It will seem lengthy, but if you can digest all of it, it should make good sense I believe.
1) First and foremost, you need to check calibration of your refractometer with every single use. That's right -- EVERY use. The readings on a refractometer really seem to bounce around a lot each time I use mine. First measure plain distilled water and ensure it reads exactly zero. If it doesn't, you can either adjust the screw in the thing (I haven't even done this) or else figure out about how far off it is from zero, then add or subtract that number of points when reading wort. For example, mine often seems to read about negative-0.2 Brix with plain water. So then I dry the thing off, then add a few drops of wort and add +0.2 Brix for an accurate reading. If you don't check calibration every time and adjust from zero, then the readings are pretty much useless and you're wasting your time. The next time you use the thing, even just 5 minutes later, check it with water again, because it probably moved! Readings are not stable, that's been my experience, so I calibrate every single time I'm going to use it. Even check it with water again immediately after measuring wort, and if the readings aren't the same in water each time, take the average of how far off it is from zero and apply that. So, for example again, if I check in plain water again and now it reads exactly 0.0 Brix, but I got negative-0.2 Brix before, figure I should add the average +0.1 Brix to the reading in wort instead of +0.2. This will provide the utmost accuracy. The more measurements and experience you gain, the more accurate your readings will be.
2) You must use both your refractometer (recording your readings IN BRIX) and hydrometer together many times with OG readings (zero ABV) along with ANY refract-to-hydro calculator (it does NOT matter which one) such as Sean Terrill's (http://seanterrill.com/2012/01/06/refractometer-calculator/
) or Petr Novotny's (see info in the OP above) so that you can learn the "correction factor" for your specific refractometer. Every refractometer is a little different in this respect. Sean Terrill suggests an average correction factor of 1.04 for many refractometers his cohorts have tested. However, do NOT assume your refractometer's correction factor is 1.04 without testing it out. Mine definitely has a correction factor of exactly 0.99. I have learned this after measuring the OG of at least 11 batches, and now that I know this correction factor, the results are always extremely consistent and predictable.
3) Okay.... so now let's say you know your correction factor. You've measured several OGs of several different worts, and it's matching up perfectly with your hydrometer reading every single time. Okay. Now you're ready to use it for FG readings! Until this point, your FG readings will be a crapshoot! But now that you know how to use the thing and apply the proper correction factor, you can use it for FG. Yay. So to do that........ I personally recommend either Petr Novotny's formulae available from the OP, or use the "Old Cubic" formula which very ironically is available in Excel spreadsheet format from Mr. Sean Terrill at: http://seanterrill.com/wp-content/uploads/2011/04/fg_calculator_v3.0.xls
Check calibration of your refractometer in water as before, then measure your final wort in BRIX, then pump that into Terrill's spreadsheet but use the result from the column that says "Old Cubic". That will give you the results you see in the chart above as red squares -- almost dead-nuts accuracy. The one goofy data point above, by the way, was where my OG was like 1.098. None of the refractometer calculators are very accurate with high gravity worts of greater than about 1.080 or 1.085. They'll still work okay for high gravity, but in my experience you'll probably need to subtract 0.003 or so to match a hydrometer reading.
Note: Hydrometer readings will ALWAYS have the greatest accuracy. Refractometers with good calculators can get you very very close indeed to a hydro reading, within 0.001-0.002, but only that accurate if you follow appropriate standard scientific practice as I've attempted to provide in all the guidance above.
I certainly hope this helps somebody more than it causes confusion. It's working really awesome for me, with my small batches (average 2 gallons). I have a lot of confidence now when using refractometers for FG and ABV that I am getting very accurate results consistent with the more reliable hydrometer readings. So now I don't need to waste a big hydro sample. Just a few drops and I'm good to go.