Compressed video stores a full picture only every few seconds (a keyframe) and cheap differences in between. A lossless cut can only begin at a keyframe, so your start point snaps to the nearest one: usually well under a second away.
Frame-accurate cutters solve the snap by re-encoding the whole clip, which is slower and spends a generation of quality. Flippy takes the lossless trade on purpose: for recordings, podcasts and highlights, the snap is invisible and the speed is not.
Audio-only files barely notice any of this: audio frames are tiny, so trims on MP3s and WAVs land almost exactly where you point.