Cosmogenic radionuclide records from polar ice cores provide unique insights into past cosmic ray flux variations. Besides allowing reconstructions of past solar activity, space weather, and geomagnetic field changes, they provide independent estimates of radiocarbon production rate changes in the past and are thus, an independent means to assess the radiocarbon calibration curve. However, all these applications rely on the proportionality of the ice core radionuclide records to the global mean production rate changes. This premise has been long debated from the model and data perspective. Here, we address this issue through atmospheric mixing model experiments and comparison to independent data. We find that all mixing scenarios that do not assume complete tropospheric mixing result in a polar bias. This bias is more prominent for geomagnetic field changes than solar modulation changes. Supported by independent geomagnetic field records and marine 10Be, the most likely scenario results in a dampening of geomagnetic field-induced changes by 23-37% and an enhancement of solar-induced changes by 7-8%. We propose a correction function that allows deconvolving the ice core to restore proportionality to the global mean signal and discuss the relevance for understanding past variations in Δ14C.