The Spaceship scene is very easy to render with a unidirectional path tracer, despite its geometric complexity. Because of this, primary-sample-space approaches are able to improve the efficiency on this scene more so than radiance-based path guiding. In fact, radiance-based path guiding worsens the error in some regions due to suboptimal sampling proportional to incident radiance on the cockpit of the spaceship. Only our product-driven neural path guiding with multiple-importance-sample-aware optimization (NPG-Product) surpasses the efficiency of primary-sample-space path sampling at an equal number of samples.