I'm not aware of any parameter that will affect circular motion in the way you have described.
Particularly given that the previous owner could not program any circular moves, there may be a problem with the actual control. I had a client some years back with a Takasawa lathe equipped with a 6TB control that could not process a circular path if the start point was not on the centre line of the arc. An arc could be programmed to finish at any angle, but if it didn't start at 3 or 12 o'clock (0 or 90 deg) then the control would raise an alarm. After I had convinced Fanuc that it was not a programming error, an eprom was replaced and that resolved the problem.
Your error is different to the above, but I would not disregard a control issue. Your program coordinates work out OK within the bounds of rounding to the least input increment, so I don't believe there is an issue with the program. Fanuc over here are very helpful, and that may be the case in your area. If you can't get a resolve, it may be beneficial to give them a call and discuss the issue.
Using R instead of I and K is a bit of a fudge in my opinion, and can result in an erroneous tool path without it being obvious if either the start or end point coordinates are wrong. In this case, and within reason, the control simply shifts the arc centre to make the circular tool path pass through the two given points. Conversely, when I and K are used, the control does a check based on the start coordinates and the arc centre described by I and K, to see if the given end point exists on the arc trajectory within a tolerance set in parameter. If it does not, an alarm is raised.