Identification in Structural Vector Autoregressions (SVARs) often relies on external proxy variables that are assumed to be valid instruments—highly correlated with a single structural shock and uncorrelated with all others. In practice, however, researchers often face a ``proxy zoo’‘ of imperfect candidates, where these exclusion restrictions are unlikely to hold. This paper develops a novel framework for set identification in SVARs that relaxes the need for valid instruments. We introduce a generalized ranking assumption, requiring only that a proxy is more strongly correlated with the target shock than with any other. This much weaker condition allows us to work with contaminated proxies that would be invalid under a standard instrumental variable approach. We combine this with traditional sign restrictions to construct sharp identified sets for monetary policy impulse responses. We characterize the geometric structure of the feasible set of structural parameters, which is formed by the intersections of spherical caps determined by the proxy information. Our method provides a robust tool for researchers to compute valid bounds on dynamic causal effects when only imperfect proxies are available.