stb results different when iprobe placed at different location of the loop

Hello, 

I ran into this issue where I got very different stb results when placing iprobe at node 1 (output of a single stage differential to single-ended OTA) as case 1, and node 2 as case 2, inside of a patented Kevin Scoones bandgap structure as attached (please ignore startup circuit on the left if they don't affect broken loop characteristics). Gain plots look not exactly the same and phase plots are drastically different, where case 1 (blue) has a down-shift of 180 deg phase and case 2 (red) has a up-shift of 180 deg indicating loop instability?)

Can someone please help me understand what have cause such a big difference?

One explanation I can find online is "The iProbe's placement can affect the impedance seen by the rest of the circuit. If the loop is broken at a point where the output impedance of a preceding stage interacts with the input impedance of the following stage, different iProbe placements can yield different results, especially at lower frequencies." but I'm not sure how it works in this particular bandgap circuit.

Thanks and regards