The precision of the result of an operation depends from the precision with which we know the data.
Example: Suppose we know the dimension of a rectangle with a precision of only one decimal digit after the point. Then, the area of the rectangle cannot have a higher precision, and hence it makes no sense to consider the second decimal digit as significant:
9.2 * 5.3 = 48.76 (the second decimal digit is not significant)
9.25 * 5.35 = 49.48 (here it is)
This is not caused by the representation of numbers in a programming language, but by the limitations on our knowledge about the input values of a problem.