Estimates of groundwater age based on 14C is often limited by the uncertainty in geochemical processes that alter the 14C concentration measured in water and the composition (δ13C and 14C) of carbon sources needed to appropriately parametrize 14C adjustment models. Estimated ages for samples that contain a mixture of young and old groundwater will be particularly sensitive to model parametrization as relatively small additions of modern 14C from recent recharge can mask the presence and amount of old groundwater. A novel multi-model approach based on inverse geochemical modeling and lumped parameter modeling of age tracers (3H, 3Hetrit, and SF6) was used to better constrain 14C dilution caused by dissolution of carbonates in the unsaturated zone or shallow parts of the Glacial aquifer, which extends over 2000 miles across the northern contiguous United States. Calibration of 14C inverse geochemical models to LPM computed 14C concentrations in modern water indicated that 14C of soil zone and shallow aquifer carbonates were not 14C-dead (0 pmC), as is typically assumed for 14C correction models. 14C of such carbonates was on average about 53 pmC (ranged 0-110 pmC, n = 72). This information was used to correct 14C concentrations for water recharged entirely before 1950 and water that is a mixture of pre- and post-1950 water. The multi-model approach developed here was compared to an analytical 14C-adjustment model (Revised Fontes and Garnier) that assumed solid carbonates were 14C-dead. 14C corrections using the analytical adjustment model tended to over-correct final 14C concentrations by 21 pmC and underestimates mean ages by 40% for groundwater mixtures. In fact, 14C corrections based on analytical model yielded negative ages (14C > 120 pmC) in nearly 36% of mixed samples. This work presents a new approach to constraining 14C corrections and age estimates of mixtures of young and old groundwater. The new method is applied to three well networks distributed across the spatially expansive Glacial aquifer.