We consider a random connection model (RCM) on a general space driven by a Poisson process whose intensity measure is scaled by a parameter $t\ge 0$. We say that the infinite clusters are deletion stable if the removal of a Poisson point cannot split a cluster in two or more infinite clusters. We prove that this stability together with a natural irreducibility assumption implies uniqueness of the infinite cluster. Conversely, if the infinite cluster is unique then this stability property holds. Several criteria for irreducibility will be established. We also study the analytic properties of expectations of functions of clusters as a function of $t$. In particular we show that the position dependent cluster density is differentiable. A significant part of this paper is devoted to the important case of a stationary marked RCM (in Euclidean space), containing the Boolean model with general compact grains and the so-called weighted RCM as special cases. In this case we establish differentiability and a convexity property of the cluster density $κ(t)$. These properties are crucial for our proof of deletion stability of the infinite clusters but are also of interest in their own right. It then follows that an irreducible stationary marked RCM can have at most one infinite cluster. This extends and unifies several results in the literature.