Others have answered the basic question well. The fundemental issue is that finite resolution binary fractions can not exactly represent decimal fractions so you always get rounding or truncation issues when using decimal fractions.