Breach and attack simulation measures control effectiveness. It answers whether the EDR blocked it, whether the rule fired, whether the SOC saw it. The question is about the defence, and the environment is yours.
The external surface raises a question that passes through no control at all. The asset answering on the internet that nobody inventoried has no EDR to block anything, because nobody knew it existed. CSURFACE starts exactly there: it discovers what answers from the root domain and validates exploitability over what it discovered.
The validation also differs in nature. It is not simulated behaviour against your stack, it is verification on the real asset that the flaw's precondition holds there and that a sequence exists leading to the described effect. When the only way to prove something would be to bring the service down, the item stays recorded as probable and never rises to proven.
Both practices coexist in a mature program. The comparison below is by capability on the external surface, which is this page's cut.