Small offsets on instrumentation are to be expected, especially on something that isn't recalibrated frequently. A couple of degrees is also not significant relative to the "idle band" used by the GS deployment logic, nor significant relevant to the discrepancy between measured TLA position and where it "should have been".
While I can't speak for AB, small discrepancies like that are pretty common - and are one reason why plots like those released/leaked, which invariably don't contain the data one would require to properly calibrate any of the parameters, must be read cautiously. When reviewing a real FDR properly, the first task is to ensure that the data has not offsets or calibration problems, and to account for them if it does. This 'advance' data may be 'raw' and not adjusted, so only gross information should be read from it until its properly validated.