Peskin & Schroeder (P&S) on p. 326 are discussing a connected, amputated 4-pt amplitude$^1$ ${\cal M}_4$. It is clear from various places in the P&S textbook [see e.g. eq. (10.21)] that ${\cal M}_4$ is an amputated amplitude. Now OP's self-loop diagram can be understood as an amputated diagram, so OP's question is still in principle a valid question.
However, P&S explain on p. 113-114 that such diagrams should be cut away. This is e.g. to avoid over-counting when we attach connected propagators $G_c$ to the external legs of the amputated amplitude ${\cal M}_4$. [Technically, self-loop diagrams in principle contribute to the connected propagator $G_c$, but are typically cancelled via renormalization conditions on the self-energy, see e.g. eqs. (10.28) & (10.29) on p. 328.]
--
$^1$ By the way, the last explicit diagram in P&S is not a 4-loop diagram: It denotes an interaction counterterm, cf. fig. 10.3 on p. 325 :-)