I have run into similar problems in the past when reading from different ADC pins. It turned out the last reading affected the current reading. The current value gets skewed higher or lower towards what the last reading was.
I solved my problem at the time by taking readings from the same pin twice, and only using the second one.
I have since learned that the problem probably comes from capacitance in the circuits. Some residual charge stays in the circuit from the previous voltage which affects the voltage on the next reading. I suspect adding capacitors to the devices being read might overcome this, but I never had a chance to test it out.