Welcome to IEEE floating point math!

Wikipedia has a good article on the specification. If you want exact fractional numbers for your own purposes, there are a number of ways around the limits of floats.

Rational will represent numbers internally as fractions, and still allow you to use the same syntax as above. The result will be "3 % 10", which you read as the fraction 3 / 10.

If you also want irrational numbers, things get tricky. CReal will do it, but has some limitations.