Most interruptions of electricity supply to final customers occur due to failures in the distribution grid. Thus, estimating its reliability is very important for planning future infrastructure upgrades. However, the estimation of reliability indexes such as SAIDI and SAIFI is highly dependent on the failure rates of different components, which are usually obtained using historical data. Despite being highly uncertain, most planning models treat failure rates as deterministic. This case study explores how failure rates uncertainty can impact the estimation of reliability in distribution systems using a model for the optimal allocation of protective devices (fuses and reclosers) and capable of estimating costs and different reliability metrics. Computational experiments on the IEEE 33-bus test system show that ignoring failure rate uncertainty can severely undermine the estimation of SAIFI and SAIDI, leading to sub-optimal decisions that do not meet minimum reliability requirements.