Quantifying seasonal bias in proxy reconstructions (for example, sea surface temperature (SST)) has been a long-standing challenge, hampering our understanding of past climate evolution (for example, the Holocene temperature conundrum). Recently, Bova et al. (Nature, 2021, 589, 548-553) proposed a seasonal to mean annual transformation (SAT) method that seems to effectively remove SST signal caused by seasonal insolation change. To extract mean annual SST (MASST) change for the Holocene epoch (12–0 thousand years before present (kyr bp)), Bova et al. selected SST records that additionally cover the last interglacial (LIG; 128–115 kyr bp) period, for which SST is assumed to be solely attributed to variations in local solar insolation, hence allowing for reliable quantification of seasonal bias in SST records. However, this assumption is fundamentally incorrect because it overlooks the roles of internal Earth system feedback (for example, sea ice) on LIG temperature change. By incorperating roles of internal feedbacks in SAT, we show that their findings are effectively biased by overcorrecting insolation-induced seasonal bias in SST proxies and thermal maxima at the last interglacial and Holocene are not seasonal origin.