A more relevant "detect-object-injection"

Is there any way that we can work towards a more helpful/relevant report of Object injection sinks?

I can't think of a relevant security use case where Object injection would be relevant outside of the scope of a function directly linked to a web service.

I can understand based on tree traversal that determining the difference in between functions that are used in response to direct network calls would be [near] impossible to determine, but if I use bracket notation at the top level of my module, likely this rule should not notify.