I have found to correct for rounding errors in decimals, to multiply each number to be summed by 100, sum the values, and then divide the result by 100. This increases the accurracy of the decimals for the summing process.