Could the 0.1uF cap across Vdd/GND have an effect here? I just swapped out the one on the breadboard for another and now the readings seem stable.