flow_net_pressure_cfm / mass_balance_ok treat a deliberately negative-pressure tent as a fault

mass_balance_ok trips whenever |intake_capacity – exhaust_capacity| exceeds max(5 CFM, 5%). A carbon-filtered grow tent is supposed to run exhaust-dominant — negative pressure is the design intent, not an imbalance. After bcb8a4b the honest numbers are intake 67.2 vs exhaust 224.4, so the flag still reads off and always will while the fans are configured this way. Separately, both entities compare capacities (nameplate x duty), never measured flow, yet publish a single hard CFM number; flow_net_pressure_cfm already labels its basis as 100pct_of_a_side_is_nameplate_proxy but the alarm built on top of it does not read that label. This is why the Critical C13 row’s acceptance line (‘the -200 alarm clears’) could not be met by calibration honesty alone — it should not have been written as a C13 criterion. Fix: either give the flag a signed target band (expected negative pressure) instead of abs(), or suppress it when either side’s basis is a proxy. That is the honesty rule applied to alarms: never raise an alarm on a basis you have already labelled untrustworthy.