Abstract: Based on about 151 years of sunspot-group observations and about 42 years of flare observations, a phase-statistical framework is constructed to test event distributions in the conjunction (\Delta\lambda=0^\circ) and opposition (\Delta\lambda=180^\circ) directions, defined by the event-planet heliocentric longitude difference rather than by inter-body alignments. To reduce observational-geometry and orbital-sampling artefacts, Earth is excluded; Cyclic Time-Shift (CTS) tests serve as the null-distribution baseline, supplemented by the false discovery rate (FDR) correction, residence-time normalization, block bootstrap, and fake-planet null tests to assess significance and robustness. The results show a one-sided conjunction enhancement for C-class flares: at the representative window half-width of 2^\circ, the conjunction ratio R_\rmC = 109.6\% (baseline 100\%; uncorrected probability p_\rmraw = 0.001; FDR q = 0.023 within the same 50-window scan family), while the opposition direction stays near the baseline. In the single-planet analysis, Venus gives the strongest signal and remains significant after correction across the seven single-planet comparisons, but cross-cycle segmentation shows that this effect is strongest in Solar Cycle 24. The full phase distribution is near-conjunction unimodal, inconsistent with the linear tidal model. No analogous robust enhancement is detected for sunspot groups as a whole, suggesting that daily sampling and solar rotation may dilute transient directional signals. Multi-planet combinations do not show a stable excess over the strongest single planet. These results support a bounded statistical association, and the underlying physical mechanism requires longer solar-cycle coverage and independent validation.