I created a large scene composed of approximately 13,000 photographs. Knowing that aligning them all at once would be unwise—as it would generate multiple components—I opted for region merging instead.
I encountered an issue during region merging. The unmerged regions contained approximately 8,600 photos with 11.9 million points in the point cloud. However, after merging the regions, the approximately 12,000 photos resulted in only 8.7 million points in the point cloud. The resulting point cloud model is extremely sparse. Previewing it in the photos reveals that many images have single-digit or even zero connected points.