Planet SkySat - Differences between image strips

I’m trying to understand why i’m seeing variations in the analytic (surface reflectance) imagery brightness values in Planet SkySat imagery between image strips that were collected around the same time.

I’m aware of the differences between analytic and visual products (i.e. this question), but i’m hoping someone can shed some light on why the analytic product contains these differences. Surely if it is being used for scientific analysis, the brightness values should be comparable between image strips? Example below (i’ve mosaicked these images into one):

Am I missing a processing step here? The visual product seems to show similar differences. My use case is image classification, so ideally I need these mosaics to contain comparable brightness values across image strips to allow me to apply a classification model consistently across the whole area.