Displacement map from A to B + Displacement map from B to C is not equal to the displacement map from A to C

I used three Sentinel-1 SLC VV radar data. I did one displacement map from (a) November 2014 to January 2015, one from (b) January 2015 to September 2015, and another from (c ) November 2014 to September 2015. When I added the raster (a) and (b) together, the values did not match that of raster (c ). Where did I go wrong?

Perhaps nowhere, the atmopheric contribution in January 2015 might be the explanation. After you remove the terrain-component of the phase, the sum of atmosphere + deformation is what is left in the interferogram.

Or maybe because I did not convert the phase to VERTICAL Displacement? I only used LOS displacement.

If both pairs (AB and BC) were acquired from the same track, this should not matter.
I agree with mengdahl that the potential contribution of atmosphere is comparably high when you only compare two images.
Assuming that the dynamics are comstant over time, you can process even more pairs to get an idea about the predominant variation of displacement.

Thank you very much, @ABraun!