In jacket-based strengthening systems, a grout layer can connect a precast ultra-high-performance concrete (UHPC) shell and an existing normal-concrete (NC) substrate, thereby forming two independent interfaces: a UHPC–grout interface and an NC–grout interface. This laboratory study evaluated the splitting-tensile and push-out responses of an untreated UHPC–grout interface and a chiseled NC–grout interface. The experimental design changes substrate type and surface treatment simultaneously; it therefore does not isolate the independent effect of chiseling. For the splitting-tensile test, the means (95% confidence interval, CI) were 3.45 ± 0.02 MPa (3.40–3.50 MPa) for SPN-Z and 3.54 ± 0.12 MPa (3.25–3.82 MPa) for SPU-G (n = 3 per group). The reported two-sided Welch test gave p = 0.324; this result does not establish equivalence. For double-face push-out testing, the complete raw PUN-Z dataset, including PUN-Z-1, gave 3.33 ± 0.38 MPa (n = 3; 95% CI: 2.40–4.27 MPa), while the untreated UHPC–grout dataset gave 1.12 ± 0.12 MPa (n = 2; 95% CI: 0.03–2.20 MPa). The push-out comparison is descriptive and exploratory only: n = 2 is insufficient for general inference, and substrate type and treatment were changed together. Under the present short-term static laboratory conditions, the specimens showed different observed failure appearances. These observations do not establish long-term, marine, durability, or cyclic-loading performance.