I have never used several swaths in one analysis, so I can’t really recommend the best procedure, sorry.
Have you tried to merge IW1 and IW2 (including the green artefacts) to check how the result looks like?
Subsetting will probably create more problems than it solves. The green artefacts happen to occur in many studies, they are not a result of false processing.
Would it make sense to use ascending instead of descending (or vice versa) so your area is fully contained within one sub-swath?
I use the snap2stamps python package to prepare S1 data for StaMPS processing
It was once announced that future versions will support the merge of different subswaths
But I don’t know when it will be released.