In NIC1, and to some extent in the other cameras, the relative sensitivity varies dramatically across the array. Regions with relatively low sensitivity will be multiplied by a large number in the calibration process, and so noise in these regions will also be amplified. This results in a modulated noise across the detector. The region in the upper left quad of NIC1 has a sensitivity that is a factor of 2-3 lower than the mean of the array, and so is most sensitive to this effect.
Complicating matters in NIC1, the linear portion of the dark current seems to be higher than the mean in this region. Thus if the incorrect dark subtraction is made then residual dark signal can remain, making this region look "higher" (or lower) than the rest. Making a correct dark subtraction will remove this positive bias, but the noise of course will still be higher due to the lower sensitivity in this region.
We haven't found a way around this one but the higher operating temperature in Cycle 11 will reduce the DQE variations.
If your object is in this region, you will have decreased signal-to-noise. The default pointing for the camera moves away from this region, but if you are unlucky in your pointing, or your target is extended and crosses this area, you will be hit by this problem.