I think the line between affricates and stop+fricative clusters is not a very sharp one. If a drunk American hallucinates an attack by bats, they may yell something about [b&tS] (modulo dialectical variance) - which sounds an awful lot like [b&t_S] <batch>. Is the onset of <tsunami>, for Anglophones who "pronounce the [t]", actually [t_s], or is it simply [ts]? I think it's clear that the /ks/ in words like "box" is a cluster rather than an affricate, since [k] and [s] have different points of articulation.