Using the method described on stack exchange here to create a shadow catcher and with Filmic render view under color space enabled, the background with shadows divided by a clean background with no shadows does not generate white. Instead, a very light grey is created where there is no shadow. This should be white so the shadow layer can be multiplied onto the composite.
However, when default render view under color space is used, the grey becomes true white, suitable to multiply with.
Therefore, I suspect there is some gotcha with exporting log color images out of blender and photoshop understanding them.
The foreground objects and the shadow layer both need to be exported to photoshop.