A detailed study of inclusive deep inelastic scattering (DIS) from mirror $A=3$ nuclei at large values of the Bjorken variable x is presented. The main purpose is to estimate the theoretical uncertainties in the extraction of the neutron DIS structure function from such nuclear measurements. On the one hand, within models in which no modification of the bound nucleon structure functions is taken into account, we have investigated the possible uncertainties arising from (i) charge symmetry breaking terms in the nucleon-nucleon interaction, (ii) finite ${Q}^{2}$ effects neglected in the Bjorken limit, (iii) the role of different prescriptions for the nucleon spectral function normalization providing baryon number conservation, and (iv) the differences between the virtual-nucleon and light-cone formalisms. Although these effects have not yet been considered in existing analyses, our conclusion is that all these effects cancel at the level of $\ensuremath{\approx}1%$ for $x\ensuremath{\lesssim}0.75,$ in overall agreement with previous findings. On the other hand, we have considered several models in which the modification of the bound nucleon structure functions is accounted for to describe the EMC effect in DIS scattering from nuclei. It turns out that within these models the cancellation of nuclear effects is expected to occur only at a level of $\ensuremath{\approx}3%,$ leading to an accuracy of $\ensuremath{\approx}12%$ in the extraction of the neutron to proton structure function ratio at $x\ensuremath{\approx}0.7--0.8.$ Another consequence of considering a broad range of models of the EMC effect is that the previously suggested iteration procedure does not improve the accuracy of the extraction of the neutron to proton structure function ratio.