Minimise floating point errors when calculating indices. Fixes #45

[3553x]

See #45 for a description of the problem. This code implements the suggested changes. I also ran benchmarks and did not observe a significant difference in terms of performance before and after the changes. All 7 tests also passed.