JSoftFloat aims to be a compliant implementation of the the IEEE 754-2008 standard Its still a major work in progress though; there are major parts of the standard which are not implemented yet. Decimal floats and logrithms have not been implemented yet.
This was initially made for use in RARS, but could certainly be useful for any applicationm where strict adherance to the standard and/or deterministic floatin point computation is needed.
If the standard disagrees with how JSoftFloat handles something, it is a bug. If you submit an issue with an example and a quote from the standard, it should be addressed in a timely manner.
Suggestions from the specification that are not followed are not bugs, but a Pull request which fufills a suggestion would be appreciated.