Drug-induced sleep endoscopy (DISE) has been progressively used to determine the individual patient responsiveness to therapy with a mandibular advancement device (MAD) for obstructive sleep apnea (OSA). This retrospective cohort study compared the general and polygraphic characteristics, as well as the sites, degrees, and patterns of upper airway collapse, in patients who responded to advancement with a titratable mandibular advancement (TMA) simulator during DISE—referred to as responders—to those in non-responders. The sample included 335 OSA patients (307 males) with a mean age of 49.98 (SD = 9.88) years, and a mean AHI of 34.14 (SD = 18.61). Once the TMA simulator customized to the patient’s dental arches was inserted and the examination was performed at 0%, 25%, 50%, and 75% of the patient’s range of antero-posterior mandibular excursion, the simulator was removed and the upper airway behavior was studied in the baseline situation. Without TMA simulator non-responders had a higher percentage of oropharyngeal complete latero-lateral and complete concentric velopharyngeal collapse. With TMA simulators, there was a significant difference between responders and non-responders in individual obstructive sites at velopharyngeal, oropharyngeal, and epiglottis levels, while at the tongue level, responders and non-responders showed the same response tendency. If confirmed in future prospective studies, these results suggest that the presence of complete latero-lateral obstruction at the oropharynx level and complete circular obstruction at the velopharynx level could be adverse phenotypes for MAD treatment outcomes in OSA patients and MAD treatment should not be considered in these patients (at least as a single therapy).