Personalizing gait-assistive robots is limited by safety and ethical constraints, which motivates studying physical human-robot interaction (pHRI) in human-in-the-loop (HITL) simulation — yet real pHRI is hard to replicate because existing models overlook the complex soft-contact dynamics that matter most during falls. We present PHRASE, a first-of-its-kind soft-pHRI simulation in MuJoCo that combines a personalized soft human model (via SKEL), an elastic strap model, and a virtual buckle–hole tightness controller. Across two subjects and three tightness levels, simulated waistline forces grow with tightening and body-mass index and reveal peak-force regions linked to discomfort, guiding more ergonomic strap and robot design.