With a redefined kilogram fixing the Planck constant, h, and a redefined mole fixing the Avogadro constant, NA, as recently proposed, retaining the carbon-12-based dalton creates a fundamental incompatibility in the stoichiometric equations. To avoid this incompatibility, the dalton should be redefined, as well: exactly in terms of the fixed-h kilogram. Specifically, with the mole defined as mol = N*/NA, where N* is an exact dimensionless constant, the dalton should be redefined as Da = (1/1000N*) kg, exactly, so that NA/mol-1 remains exactly equal to the gram-to-dalton mass-unit ratio—as required by the fundamental compatibility condition relating the kilogram, mole and dalton. This would necessarily decouple the dalton from the carbon-12 reference mass, ma(12C)/12, used for cataloguing nuclidic mass ratios to extremely high precision. Relative atomic masses, defined as Ar(X) = ma(X)/Da, would then have uncertainties dominated by that of ma(12C)/kg; but since this would be of order 10-9 or less, the values of relative atomic masses, to the precision used in stoichiometry, would be totally unaffected. This is a much more straightforward and easily comprehended strategy than those using confusing inexact correction factors or the equivalent, as recommended in recent publications and implicitly endorsed by the CGPM.