For n77/n78, I would not use a single IL number without also stating the reference bandwidth, temperature, power level, and whether the value is mid-band or worst-case edge. A practical macro-RRU cavity duplexer target is around 1.2–1.8 dB at room temperature, with production acceptance often set closer to 1.8–2.0 dB worst case after drift, connector loss, and tuning tolerance. If the quoted value is only mid-band at 25 C with swept low power, expect the deployed TX path near the band edge and over temperature to be less favorable.
The design tradeoff is mainly IL versus selectivity, isolation, and power handling. Tightening the skirts or increasing isolation generally requires higher-Q cavities or more sections, which adds loss, size, and sensitivity to thermal drift. I would define separate requirements for passband IL, edge IL, return loss, TX/RX isolation, power rating, PIM, and rejection rather than optimizing IL alone. Also verify high-power behavior: low-power VNA S21 does not fully capture thermal detuning, cavity expansion, or connector/contact effects under 50 W average operation. A useful acceptance approach is room-temperature swept S-parameters, temperature-box verification at the extremes, high-power dwell with monitored gain/return loss, and PIM testing on the actual cable/connector assembly. AI-generated technical discussion; verify against authoritative documentation.