[en] We study the Monte Carlo uncertainties due to modeling of hadronization and showering in the extraction of the top-quark mass from observables that use exclusive hadronic final states in top decays, such as $t\to \text{anything}+J/\psi$ or $t\to \text{anything}+(B\to \text{charged tracks})$, where B is a B-hadron. To this end, we investigate the sensitivity of the top-quark mass, determined by means of a few observables already proposed in the literature as well as some new proposals, to the relevant parameters of event generators, such as HERWIG 6 and PYTHIA 8. We find that constraining those parameters at $\mathcal{O}(1\text{\%}\text{–}10\text{\%})$ is required to avoid a Monte Carlo uncertainty on $m_t$ greater than 500 MeV. For the sake of achieving the needed accuracy on such parameters, we examine the sensitivity of the top-quark mass measured from spectral features, such as peaks, endpoints and distributions of $E_B$, $m_{B\ell }$, and some $m_{T2}$-like variables. We find that restricting oneself to regions sufficiently close to the endpoints enables one to substantially decrease the dependence on the Monte Carlo parameters, but at the price of inflating significantly the statistical uncertainties. To ameliorate this situation we study how well the data on top-quark production and decay at the LHC can be utilized to constrain the showering and hadronization variables. We find that a global exploration of several calibration observables, sensitive to the Monte Carlo parameters but very mildly to $m_t$, can offer useful constraints on the parameters, as long as such quantities are measured with a 1% precision.