I'm attempting to do a spatial join and unfortunately the 2 input datasets have different spatial references. This means that the 2nd is transformed to match the first. The data should align, but the transformation means there are slight differences. These differences are resulting in features joining that should not when using the "intersects" or "contains" options. In ArcGIS Pro, i frequently add a negative search distance to reduce the chance of overlaps around the edges, but this is not possible (yet) in Data Pipelines.
I even tried the "Near geodesic" option with a negative distance (5 metres) but that results in an error when run/previewed.
In the screenshot below, you can see the incorrectly join areas circled. The blue hatch indicates what should have been joined - the solid yellow areas have only joined based on the slightest overlap/touch at the boundaries.
I even tried to include a projection step prior to the spatial join to remove the slight shift, but it made no discernable difference.