The effectiveness of a universal adhesive applied in three application modes for the preparation of Class V composite restorations was evaluated both clinically and by quantitative marginal analysis (QMA) over 36 months. In 50 patients, three (n = 21) or four (n = 29) non-carious cervical lesions (NCCL) were restored with Venus® Diamond Flow (Kulzer GmbH, Hanau, Germany). The adhesive iBond® Universal (iBU, Kulzer, Germany) was used in self-etch (SE), etch-and-rinse (ER), or selective-enamel-etch mode (SEE). The etch-and-rinse adhesive OptiBondTM FL served as a control (OFL, Kerr GmbH, Herzogenrath, Germany). The restorations were clinically assessed (FDI criteria) at 14 days (BL), 6, 12, 24, and 36 months. Additionally, QMA was conducted on all restorations of 11 randomly selected patients. FDI criteria and marginal gap and perfect margin were compared between and within groups and recalls using McNemar, Wilcoxon, or Mann-Whitney U-tests (α = 0.05). Starting with 12 months, cumulative failure rates were lower in iBU-SE (0.0%, p = 0.016) and iBU-ER groups (2.1%, p = 0.07) compared to OFL (16.7%). At two years, iBU-SEE also showed fewer failures (0.0% SEE vs. 34.6% OFL, p = 0.016), as did iBU-SE compared to iBU-ER after 36 months (2.2 and 19.6%, p = 0.039). From BL, the iBU-SEE group always had the fewest marginal gaps and the highest percentage of perfect margins. From BL, iBU-SEE (0%, p = 0.008) and iBU-ER (0.2%, p = 0.027) showed significantly fewer marginal gaps compared to OFL (2.5%) and more perfect margins were found with iBU-SEE starting at 6 months (p = 0.054). The SEE and ER modes ensured the most excellent marginal quality, with differences from the control appearing earlier with QMA than clinically. In restoring NCCls, iBU showed superior clinical performance over OFL, especially in modes SE and SEE.