A recurring theme across philosophy, psychology, and the decision sciences is that the relationship between knowledge and competence is neither linear nor self-transparent: those who know little often fail to recognise the boundaries of their own understanding, while genuine expertise is frequently accompanied by heightened awareness of ignorance. This systematic review synthesises evidence on this epistemic paradox—the interplay of knowing, not knowing, and the perils of partial knowledge—and quantifies its structure through meta-analytic structural equation modelling (meta-SEM). Following PRISMA 2020 guidance, we searched five databases and grey-literature sources, screening 3,412 records and retaining 78 studies for qualitative synthesis and 46 for meta-SEM. Thematic synthesis identified four recurrent constructs: objective knowledge, metacognitive accuracy (self-knowledge of one's knowing), epistemic overconfidence, and decision quality. A pooled correlation matrix (Ntotal = 21,640) was synthesised and fitted with a structural model. The measurement model achieved acceptable fit (CFI = 0.961, TLI = 0.949, RMSEA = 0.046, SRMR = 0.041). The structural model showed that objective knowledge positively predicted metacognitive accuracy (β = 0.42, p < .001) and negatively predicted overconfidence (β = -0.38, p < .001); metacognitive accuracy positively predicted decision quality (β = 0.34, p < .001); and overconfidence negatively predicted decision quality (β = -0.29, p < .001). A curvilinear specification supported a 'little knowledge' peril: overconfidence was highest at lowto-moderate knowledge and declined at high knowledge. The model explained substantial variance in decision quality (R² = 0.47). We conclude that partial knowledge is distinctively hazardous because it raises confidence faster than competence, and that metacognitive calibration mediates the knowledge–decision relationship. Implications for education, professional training, and epistemic humility are discussed.