Graft versus host disease (GVHD) is the major limitation of allogeneic hematopoietic stem cell transplantation (HSCT) presenting high mortality and morbidity rates. However, the exact cause of death is not completely understood and does not correlate with specific clinical and histological parameters of disease. Here we show, by using a semi-allogeneic mouse model of GVHD, that mortality and morbidity can be experimentally separated. We injected bone marrow-derived dendritic cells (BMDC) from NOD2/CARD15-deficient donors into semi-allogeneic irradiated chimaeras and observed that recipients were protected from death. However, no protection was observed regarding clinical or pathological scores up to 20 days after transplantation. Protection from death was associated with decreased bacterial translocation, faster hematologic recovery and epithelial integrity maintenance despite mononuclear infiltration at day 20 post-GVHD induction with no skew towards different T helper phenotypes. The protected mice recovered from aGVHD and progressively reached scores compatible with healthy animals. Altogether, our data indicate that severity and mortality can be separate events providing a model to study transplant-related mortality.