Tests run on interpolation algorithms applied to grids of subgiant models with typical resolution have revealed the algorithms' inadequacy in rendering l=1 mode frequencies within the precision expected from long-term space-based observations (~0.1 μHz). Interpolation errors are particularly concerning in the case of mass tracks associated with the transition between radiative and convective cores during the main sequence. In such cases, they show systematic errors throughout the subgiant phase, reaching maxima of 7.5 μHz. Here, we show that these large systematic errors stem from the drastic change in the core structure in adjacent mass tracks when the interface between convective cores and radiative interiors during the main sequence is modelled by means of exponential diffusive overshooting with convective overshoot parameter, f_ov. Moreover, we discuss how f_ov influences interpolation in a grid of subgiant models and how the issues caused by less educated guesses of its value can be mitigated. Additionally, we explore the possibility of creating a grid of subgiant models where f_ov is dependent on the mass, as discussed in Claret & Torres (2018), and highlight the benefits associated with this approach.