Stack mystery

Hello all - I am having a stack issue that I have been unable to resolve. I am attempting to look at backscatter statistics (mean, standard dev, etc.) by pixel for 3 years of Sentinel 1 SLC data (2015-2017). However, some images are misaligned in the final stack. I have spent some time trying to resolve this error. I am hoping for some advise on what might be causing this error and how to fix it.

I am using the command line interface, My workflow is:
step 1) apply TOPSAR-Split to each scene, selecting bursts by a polygon
step 2) apply a graph with does the following: apply orbits, remove thermal noise, calibrate to sigma0, and deburst
step 3) apply Subset to each scene, clipping to the same polygon in step 1
step 4) apply the Radar->Standard Graphs->Coregester.xml graph with default options, except for “resampling type”, which is set to “bilinear_interpolation”

This is a clip of two scenes in radar coordinates showing backscatter converted to db. The orange areas are lakes. The lakes are shifted about 15 pixels in the azmouth.

Many thanks for your help with this.
Kim