A JPG encoder looks for detail the eye is less likely to notice and stores an approximation. Quality is not a percentage of the original bytes. It is a setting the encoder interprets.
Canvas export in the browser also strips most metadata, including GPS coordinates, because it draws pixels onto a new bitmap. That is useful, and it is not the same as a careful metadata editor.
If the output is larger, the original was probably already compressed, or the new format is a poor match. The compressor on this site reports that honestly.