The size of the duplicate instance is worked out by the modifier reference objects scale(in the example the empty) divided by the active/modified objects scale (in the example the step) times the previous objects size. So at the point in the video you’re referring to the empty was set to scale of 1 – so when increasing the object to 1.215 the duplicates where 1 / 1.215 of the previous, meaning they got smaller. This is why keeping the original object scale set to 1 makes it easier as then there is no division to complicate things and the changes in size come directly from the empty scale values.
