sorry I thought this question was already answered but it was just your added screenshot - the opposite direction approach always requires changes on two sides, for one on the process that has waste output (move waste to input, change sign to negative) and second for the waste treatment process, move waste that is input to output and change the sign there as well. I call it therefore a "double negative" approach. If you forget to change the sign on one of these sides, you get negative results obviously for this life cycle branch. The waste treatment processes are modeled by yourself, right (I did not check now but sPET use does not sound like an ecoinvent process) - then this could be the issue.
You can of course add pictures in posts, you can also edit your post to avoid adding an additional answer or new post btw..