The infrared (IR) stimulated luminescence (IRSL) and post-IR IRSL (pIRIR) signals from K-feldspar can, for convenience, be divided into two components, bleachable and ‘non-bleachable’, where the latter corresponds to the ‘residual’ signal observed in sunlight-bleached samples. In this paper, we examine the non-bleachable component of IRSL of K-feldspar for several sedimentary samples from across Eurasia. We observed a large variability in the residual doses among these samples after prolonged exposure to sunlight. By employing multiple elevated temperature (MET) IR stimulations at 50–300 °C, we show that the residual dose increases systematically with stimulation temperature, attaining values as high as ∼50 Gy at 300 °C, even after several hours to tens of hours of exposure to unfiltered sunlight. We examined two samples in detail and found that the bleachable and non-bleachable components produced different dose response curves. Pulse annealing studies showed that the non-bleachable component is more stable than the bleachable component, suggesting that a preheat procedure cannot eliminate the non-bleachable component. Additional experiments revealed that the non-bleachable component is dose dependent. Owing to this dose dependency, we demonstrate mathematically and empirically that the simple subtraction of a residual dose from the measured equivalent dose (De) – which is the most common approach employed (if any residual dose is subtracted at all) – will result in underestimation of the actual De. We present a method to correct for the dose dependency of the residual dose, which can improve the accuracy of either MET-pIRIR or pIRIR age estimates for samples in which the non-bleachable component represents a significant fraction of the measured signals.