URB Paper #403: Consciousness in Uploaded Minds — Part II: Discrete IIT-Φ, Global Workspace Lesion, φ-Scaling, and the Generalized Mirror Test

URB #402 identified two methodological failures: the Gaussian IIT-Φ approximation collapses for sparse spike trains, and the Mirror Self-Recognition protocol was underspecified. This paper corrects both. Four new tests are applied to the C. elegans touch circuit (6-neuron LIF) and an extended 12-neuron model (touch + thermotaxis + chemotaxis): (1) Discrete IIT-Φ using exact spike-pattern entropy, (2) Global Workspace lesion study — identify the bottleneck neuron by systematic removal, (3) φ-scaling of post-stimulus decay, (4) Generalized MSR via LCC self-other cross-correlation across 20 trials. Combined score across URB #402 and #403: **7/13 unique criteria met**. The single most statistically robust finding: the Generalized MSR is strongly positive (t=5.882, p<0.0001, Cohen's d=1.907) — the network reliably distinguishes its own dynamics from another network's across 20 randomized trials. The failures of Φ and φ-scaling are shown to reflect a specific architectural property — **the reflex arc is not the locus of consciousness** — and point toward the multi-modal integration layer as the candidate substrate.