Hi Donatas,

as John describes, your result was absolutely correct.

And on most floating point libraries the proper rounding can be achieved with the FLOATFUZZY directive.
E.g. on the Keil C51:

http://www.keil.com/support/docs/1491.htm

Peter