The conversation after a flagged submission is usually worse than it needs to be, because both sides are defensive. The student is afraid of an accusation, the teacher is afraid of being fooled, and the report is treated as a verdict rather than a signal.
What works is boring. Say what you actually used, for what part of the process, and bring the trail: notes, the outline, an earlier draft, the version history. If a tool rewrote sentences, say that, and explain that you edited the result. A concrete process is more persuasive than any argument about the accuracy of detectors.
Before that conversation exists, the evidence is easy to keep: write the outline yourself, keep one draft file, and note which tools touched which paragraphs. That is a two-minute habit.
This explainer on Turnitin's similarity colours helps separate the two numbers in the report, and this page on being wrongly flagged is a sensible preparation read.