High-resolution regional climate projections are essential for impact assessments and adaptation studies in Japan, where the dynamically downscaled 5‑km d4PDF dataset and the statistically downscaled 1‑km NIES2019/NIES2020 datasets are widely used. Despite their central role in national climate‑impact research, the consistency of future climate change signals between these major downscaling products has not been systematically evaluated. The primary objective of this study is to identify inter-dataset discrepancies in projected climate change over Japan. To this end, we conduct a multi-variable comparison of future climate projections from the two datasets. Projections from NIES2019/NIES2020 generally follow their driving global climate models, showing increases in temperature, precipitation, surface solar radiation, and downward longwave radiation, along with decreases in surface wind speed and relative humidity as global mean temperature rises. In contrast, d4PDF shows a suppressed increase in precipitation and a decrease in surface solar radiation. Comparisons with multiple global climate models and previous studies of European and North American climate change suggest that these discrepancies partly arise from the absence of future reductions in anthropogenic aerosol emissions over East Asia within the d4PDF dynamical downscaling framework. This omission suppresses the expected brightening and consequently dampens precipitation increases. Overall, our results reveal substantial inconsistencies among major Japanese downscaling products for several key variables, highlighting the need for careful dataset selection when conducting climate‑impact and adaptation assessments for Japan.