Its the "around 2.5 Volts" that is of concern. Still a 0.07 volt disparity is excessive.
I would:
- Print out the raw A/D counts for both A0 and A1.
Print them for more that a few seconds to see if the readings are stable.
It would be best to use the IDE Serial Monitor if you can. (simple is best for troubleshooting) - with your voltmeter measure the A0 and A1 voltage right at the board.
- Also measure with your voltmeter the 5V right at the board.
- Don't change the reference (for now).
These numbers should allow the error source to be found.