Not exactly what you are probably wanting to hear, but, you'd honestly probably be ahead, at least accuracy wise, to start at a known surveyed point and simply use relative distance measurements (example: 500 ft from survey marker 90 degrees - due east). I know some groups who have used a method such as that using a digital lazer range finder device (had built in compass as well, along with bluetooth to make recording data easy). They initially tried it using sub-meter professional GPS units, but they had too many trees in the way in their area to get the accuracy you need for grave sites even with those GPS units.
All that being said, your suggestion of using a survey point as a control would be better than nothing. If you are going to GPS each individual grave site like that though, I would make sure that you pay attention to the day you go out (minimize any bad weather or possible interference such as that) and I would make sure to not just take a single quick GPS reading, especially on the control point. One of the ways many professional GPS systems improve the accuracy of their readings is by not actually taking one reading per recorded point, but rather they may take anywhere from 10 to 30 or more readings (may take a couple of seconds to a couple of minutes depending on system and settings) and average the locations together to get a more accurate location that compensates for any momentary interference. There are also systems out there for professional GPS units/software that actually perform the same kind of correction you are talking about here, but using known regional survey control points that publish their correction measurements online, but I'm personally not aware of that really being usable with standard generic cell phones without a fair bit of specialized software... though if someone else knows of a good solution for this, please chime in.
Just my thoughts, hope they help.