A Cautionary Tale of Mis-measured Tails from $q/g$ Bias

Jet substructure techniques such as subjet $p_T$-asymmetry, mass-drop, and grooming have become powerful and widely used tools in experimental searches at the LHC. While these tools provide much-desired handles to separate signal from background, they can introduce unexpected mass scales into the analysis. These scales may be misinterpreted as excesses if these are not correctly incorporated into background modeling. As an example, we study the ATLAS hadronic di-$W/Z$ resonance search. There, we find that the substructure analysis -- in particular the combination of a subjet asymmetry cut with the requirement on the number of tracks within a jet -- induces a mass scale where the dominant partonic subprocess in the background changes from $pp \to g \!+\! q/\bar q$ to $pp \to q\bar{q}$. In light of this scale, modeling the QCD background using a simple smooth function with monotonically decreasing slope appears insufficient.