ABSTRACTOne-zone Galactic chemical evolution (GCE) models have provided useful insights on a great wealth of average abundance patterns in many environments, especially for the Milky Way and its satellites. However, the scatter of such abundance patterns is still a challenging aspect to reproduce. The leading hypothesis is that dynamics is a likely major source of the dispersion. In this work, we test another hypothesis, namely, that different assumptions on yield modelling may be at play simultaneously. We compare whether the abundance patterns spanned by the models are consistent with those observed in Galactic data. First, we test the performance of recent yield tabulations, and show which of these tabulations best fit Galactic stellar abundances. We then group the models and test if yield combinations match the data scatter and standard deviation. On a fixed Milky Way-like parametrization of NuPyCEE, we test a selection of yields for the three dominant yield sets: low-to-intermediate mass stars, massive stars, and Type Ia supernovae. We also include the production of r-process elements by neutron star mergers. We explore the statistical properties spanned by such yields. We identify the differences and commonalities among yield sets. We define criteria that estimate whether an element is in agreement with the data, or if the model overestimates or underestimates it in various redshift bins. While it is true that yields are a major source of uncertainty in GCE models, the scatter of abundances in stellar spectra cannot be explained by a simple averaging of runs across yield prescriptions.