Han et al. (2025) examine Chinese medical researchers with retracted publications, showing that career penalties are sharply stratified by seniority. Early-career researchers face lasting declines in citations, collaboration, and career mobility, whereas senior researchers remain largely insulated, reflecting institutional and cultural leniency. The study further shows that output-driven incentives increase retraction risk, broad collaboration networks mitigate it, and retractions due to scientific errors can result in harsher penalties than those from misconduct. Despite its rigor, the study stops short of critiquing the retraction system. Treating all retractions as equivalent undermines fairness and scientific integrity by conflating deliberate misconduct with honest error, disproportionately harming early-career researchers and discouraging self-correction. Drawing on principles of proportionality and intent from legal frameworks, we propose a principled retraction typology distinguishing malicious intent, negligence, and honest mistakes to preserve accountability while safeguarding the self-correcting function of science. We further argue that Scopus and Web of Science exacerbate these harms by applying a single undifferentiated retraction label, amplifying stigma and career damage. Differentiating retractions on these platforms is urgent to reduce unjust penalties, promote transparency, and align bibliometric practices with the self-correcting norms essential to scientific progress.