Geostationary optical lightning imagers provide continuous observations of total lightning activity, but their response to upward lightning and continuing-current-dominated flashes remains difficult to evaluate because direct current-based ground truth is rarely available. This paper evaluates Meteosat-12 Lightning Imager (LI) observations using direct lightning-current measurements from an instrumented wind turbine in Croatia. The turbine is equipped with a current-measurement system, and selected events are additionally documented using high-speed video and observations from a ground-based lightning location system. We analyze 76 direct strikes to the instrumented wind turbine during the LI pre and operational period from July 2024 to February 2026. The dataset consists predominantly of upward flashes, with only five downward cloud-to-ground cases, and includes negative, positive, and bipolar flashes. It also encompasses waveform classes dominated by initial continuous current (ICC), as well as flashes containing pulse-type components such as return strokes (RSs). For each measured lightning current waveform, LI group-level observations were extracted within 50 km of the turbine and referenced to the onset and duration of the measured current. LI detections were evaluated in a preceding window, from 500 ms before current onset to the beginning of current flow, and in a during-current window extending over the measured current duration. The overall LI detection efficiency was 76.3 % when detections in either window were considered, with efficiencies of 61.8 % in the preceding window and 65.8 % during the measured current flow. LI detectability depended on current waveform class, with flashes containing pulse-type components generally more frequently detected during current flow. Furthermore, LI detected groups more frequently in the preceding window for positive flashes initiated by negative upward leaders than for their negative-polarity counterparts. In contrast, the detection efficiency during the current-flow window was the same for both polarities. The day/night analysis showed a modestly higher detection efficiency at night. Moreover, group-count, radiance, and flash-level analyses show that LI observations mainly represent the cloud-top optical manifestation of the broader lightning process rather than the local attachment process alone. These results demonstrate the value of direct lightning-current measurements for evaluating satellite optical lightning observations of lightning.