Is patched code used to generate bytecode?

I've been using error prone to fail the build, for example using `-Xep:BooleanParameter:ERROR`
Then I come along and patch the code before rebuilding.
But now I want it to patch the code as we go `-XepPatchLocation:IN_PLACE -XepPatchChecks:BooleanParameter`

My question is, is the patched code or the original code used to generate the bytecode? And perhaps this could be mentioned at https://errorprone.info/docs/patching

What's confusing is that if I leave the ERROR check in together with the patch, the code gets patched. But then rather than getting an ERROR for each violation, or no log at all, I get WARNINGs. So its as if the check is considering the original code, and afterward the severity of the results are reduced by the patch. Fine. But is the patched code used to generate the bytecode?

Does the same order of operations apply with refaster and inlineme?