Tissue constructs containing mesenchymal stem cells (MSCs) are appealing strategies for repairing large segmental bone defects, but they do not allow consistent bone healing and early cell death was identified as a cause of failure. However, little is known about cell survival in the clinical microenvironment encountered during bone healing process. Osteoconductive coral scaffold with or without luciferase-labeled human MSCs were implanted either in a critical segmental femoral bone defect stabilized by plate or subcutaneously in 44 mice. Cell survival was evaluated by serial bioluminescence imaging (BLI) and osteogenic capabilities by histology and microcomputed tomography. Comparisons between groups were performed with two-way analysis of variance test. Twenty mice were sacrificed 2 weeks after surgery for short-term evaluation and 24 mice at 10 weeks for long-term evaluation. BLI provided evidence of fast and continuous cell death: 85% decrease of the BLI signal over the first 2 weeks in both locations; in fact, less than 2% of the initial cell number was present in all constructs analyzed 4 weeks postimplantation and less than 1% of the initial cell number by 8 weeks postimplantation. By 2 weeks postimplantation, the amount of newly formed bone was self-limited and was similar to ectopic and orthotopic groups. By 10 weeks postimplantation, bone formation was significantly enhanced in the presence of MSCs in orthotopic site and the amount of newly formed bone in cell-containing constructs implanted in orthotopic locations was significantly higher than that observed in the ectopic group. Our results indicated that hMSCs promote bone formation despite early and massive cell death when loaded on coral scaffolds. Interestingly, bone formation was higher in orthotopic than ectopic site despite the same survival pattern. Ectopic implantation of cell-containing constructs is suitable to evaluate cell survival, but assessment of bone formation ability requires orthotopic implantation.