Cameras and phones record at bitrates far above what watching requires, because they must encode in real time with no second look at the footage. A re-encode gets to study the video properly and spend bits only where the picture needs them: that gap is your 30 to 70%.
Screen recordings compress hardest of all. Most of the frame barely changes from one moment to the next, and modern codecs are excellent at not paying twice for the same pixels.
A file that has already been compressed twice has little left to give: each pass spends the headroom. If your source is a downloaded, re-shared copy of a copy, expect modest savings and be gentle with the preset.