Precise manipulation, fast interaction, and lasting comfort are hard to reconcile in VR. Modern controllers can track at sub-millimetre resolution in ideal conditions, yet mid-air reach, tremor, and arm elevation still impair accuracy and induce fatigue. We present Eyes on Target, Pen on Table, combining gaze-driven target selection with desk-supported, pressure-adaptive pen input. In a within-subjects study (N=12), it was compared with a direct mid-air pen baseline. In 2D tasks, mean placement error dropped by 40% with no significant time cost, while in 3D tasks error fell by 58% at a 52% time cost. With control-display gain with axis constraint, error reductions reached 77% in 2D and 81% in 3D relative to the baseline. Across conditions, the approach also reduced arm fatigue and NASA-TLX workload. Our findings position the Gaze + Pen on Table technique as a balanced, sustainable alternative for production-level VR design, with practical guidelines for implementation.