Even though it’s a wireframe view, the first image in grayscale appears blurry or as though the elevations are squelched. Whereas in the actual map (2nd image), they’re more distinguishable and have a better gradient transition representing the terrain. It could be the Scale values are too high (X=2500, X=2500, z=15000), so the results in Unreal of interpolating based on the quads in components is watering down the distinctions in height. Basically it’s modifying the slope to be noticeably more gradual than the actual map.
2500 is 25 meters per quad, and there’s 512 quads / component in 2x2 sections, so 512x2 = 1024. Thus, 25,600 meters, which I’m thinking is the map’s coverage area, correct? or is it smaller than that? Decreasing the scale of X and Y, then increasing the number of components proportionally to get the same coverage area may be a part of the solution. And if you weren’t using 16-bit before, it would also explain the flattened result in Unreal. Interesting though, not using 16-bit could also produce exaggerated height / slope representation in other extreme too, with steeper slopes and sharp terrain transitions.
150 meters per Z axis sounds rather tall for the map in 2nd image. What are the lowest and highest points in the original heightmap?