Difference between packaged and standalone for a postprocess material reading world normal and camera vector

I have a post process material i am using for a simulation sensor, with a scene capture component. However it’s giving different results for packaged builds vs running in standalone, off by 1 2 or 4/4096 at each pixel . I think the standalone performance is what I want but I can’t figure out a way to get packaged to match.

how do I make it match?