Pioneer vs RPL precision

[Marco Polo]

Yesterday I was playing with EMU42 and realized that is some circumstances Pioneer calculators (no matter if RPN or algebraic) shows better precision than RPL models (from 28s to 50g).

The operation is the classic "(10/3)*3".

On any Pioneer and even 35s the result is always 10, even if calculated as a program or an algebraic expression

On any RPL model the result is:
- 9.9999999999 if calculated as "10 <ENTER> 3 / 3 *"
- 9.9999999999 if calculated as << 10 3 / 3 *>>
- 9.9999999999 if calculated as '10/3*3" <EVAL>
- 10 if the operation order is changed, i.e. '3*10/3', regardless the way it is calculated
The reason of RPL models behaviour appears logic to me, as the result of 10/3 is approximated and the error is propagated to the next operation. In '3*10/3" there is no approximation.

It is not clear to me how Pioneer models handles the approximation much better.