Dose escalation during early phase clinical trials is a difficult balancing act between trying to reach efficacious dose levels quickly and avoiding excess toxicity. To guide this process, studies use statistical models, such as the popular Bayesian Logistic Regression Model (BLRM), often utilizing external data. The external data is incorporated into the prior for BLRM, but a highly informative prior risks poor dose selection if the prior is misspecified. Utilizing a mixture prior that combines a highly informative prior with a weakly informative prior mitigates that risk, but it also adds another parameter that needs to be specified for the model. We provide guidance for incorporating mixture priors into BLRM using both simulation and visualization tools. These tools help integrate informative priors into dose escalation while avoiding an inflexible prior, thus safeguarding BLRM against prior misspecification.