25% speed at 60fps means 15fps effective output, which is well under the human eye’s ability to detect motion. An effective 15fps is always going to look stuttery because the eye is fast enough to detect individual frames at that speed. For 25% slowdown to work, the source would need to be 100-120fps, which would result in 25-30fps output after being slowed down. Also, at the time of capture, the motion blur (shutter speed) needs to be calculated for the slow version of the video, not the real-time frame rate.
Shotcut repeats frames in order to stretch a video from real-time to slow-motion. In the case of your project, the input is 60fps and the output is 60fps. The input doesn’t have any “extra” frames to fill time when the clip is slowed down, so Shotcut repeats frames to fill the extended time.
There is a technique called motion interpolation (or optical flow) which attempts to mathematically construct in-between frames when doing a slow-down. Shotcut does not support this natively, but the footage could be prepared outside of Shotcut using FFmpeg if you’re adventurous. If going this route, the footage needs to be captured with as fast a shutter speed as possible to create blur-free edges so the algorithm will have a higher chance of matching edges between neighboring frames. See these links: