The proton, deuteron and triton masses can be determined relative to the electron mass via rovibrational spectroscopy of molecular hydrogen ions. This has to occur via comparison of the experimentally measured transition frequencies and the ab initio calculated frequencies, whose dependence on the mass ratios can be calculated precisely. In precision experiments to date (on HD$^+$ and H$_2^+$), the transitions have involved the ground vibrational level $v=0$ and excited vibrational levels with quantum numbers up to $v'=9$. For these transitions, the sensitivity of the ab initio frequency on the high-order-QED contributions is correlated with that on the mass ratios. This prevents an efficient simultaneous determination of these quantities from experimental data, so that the accuracy of the mass ratios is essentially limited by the theoretical uncertainty. Here we analyze how the accuracy of mass ratios may be improved by providing experimental transition frequencies between levels with larger quantum numbers, whose sensitivity on the mass ratio is positive rather than negative, or close to zero. This allows the unknown QED contributions and involved fundamental constants to be much more efficiently determined from a joint analysis of several measurements. We also consider scenarios where transitions of D$_2^+$ are included. We find these to be powerful approaches, allowing in principle to reach uncertainties for the mass ratios two orders smaller than CODATA 2018, and without using muonic hydrogen data. For the Rydberg constant and the charge radii, improvements by factors of 4 to 11 are projected.