Geoscientists infer past plate motions, which serve as fundamental constraints for a range of studies, from observations of magnetic isochrons as well as hotspots tracks on the ocean floor and, for stages older than the Cretaceous, from paleomagnetic data. These observations effectively represent time-integrals of past plate motions but, because they are made at present, yield plate kinematics naturally tied to a present-day reference-frame, which may be another plate or a hotspots system. These kinematics are therefore different than those occurred at the time when the rocks acquired their magnetisation or when hotspot-related marine volcanism took place, and are normally corrected for the reference-frame absolute motion (RFAM) that occurred since then. The impact of true-polar-wander events on paleomagnetic data and the challenge of inferring hotspot drifts result in RFAMs being less resolved – in a temporal sense – and prone to noise. This limitation is commonly perceived to hamper the correction of plate kinematic reconstructions for RFAMs, but the extent to which this may be the case has not been explored. Here we assess the impact of uncertain RFAMs on kinematic reconstructions using synthetic models of plate motions over 100 million years. We use randomly-drawn models for the kinematics of two plates separated by a spreading ridge to generate a synthetic magnetisation pattern of the ocean floor. The kinematics we infer from such a pattern are outputs that we correct for synthetic RFAMs using two equivalent methods (a classical one as well as another that we propose and test here) and then compare to the ‘true’ motions input. We assess the misfits between true and inferred kinematics by exploring a statistically-significant number of models where we systematically downgrade the temporal resolution of RFAM synthetic data and add noise to them. We show that even poorly-resolved, noisy RFAMs are sufficient to retrieve reliable plate kinematic reconstructions. For relative (i.e., one plate with respect to another) and absolute (i.e., relative to the deep mantle) plate motions, estimates upon RFAM correction differ from the true kinematics by less that 10% and 3%, respectively.