Disordered pinning models are statistical mechanics models built on discrete renewal processes: renewal epochs in this context are called contacts. It is well known that pinning models can undergo a localization/delocalization phase transition: in the localized phase the typical density of contacts is positive and the largest gap between contacts is at most of the order of the logarithm of the size $n$ of the system, whereas the system is void of contacts in the delocalized phase. When disorder is absent and the phase transition is discontinuous, conditioning the contact density to be positive but smaller than the minimum typical density in the localized phase has the effect of forcing to create one, and only one, macroscopic gap between two contacts, while the rest of the configuration keeps the characteristics of a localized state. However it is known that, in the presence of (bounded) disorder, this big jump phenomenon is no longer observed, in the sense that the largest gap in the conditioned system is $o(n)$. Under minimal integrability conditions on the disorder we show that the conditioned system is localized in a very strong sense, in particular the largest gap is $O(\log n)$. The proof is achieved by exploiting the improved understanding of the localized phase of disordered pinning models developed in [34] and by establishing some refined estimates, in particular a quenched Local Central Limit Theorem. We also present an analysis of the effect of disorder on the large deviation rate function of the contact density and, as an important ingredient for the generality of our results, we establish a smoothing inequality for pinning models under minimal integrability conditions, thus generalizing [11,32].