You have to think in three dimensional terms of the aircraft rotating around the gyroscope.
As was pointed out, the instrument's max error occurs at the 180 degree mark of a turn. That's because for the first 180 degrees of a circle, the instrument is recalibrating itself based on the G forces. It is being turned on its side.
However, remember how the gyro is "stuck in space" and the fuselage is pivoting around it? Well, just as there was a force for the first half of the 360, pushing the gyro to the side, during the second half of the 360 the same force will be pushing the gyro back towards its original position.
Does that help at all? Clear as mud?