With no e3000 handy and having taken very little time to analyze your post,
I'll stick my neck out and risk embarassing myself with the following
response to your post:

I'd suggest that the problem is perhaps simply one of imprecision.  Real
numbers are stored as binary characteristic and mantissa.  That makes it
impossible to represent certain numbers exactly.  Input and output of real
numbers is subject, therefore, to errors due to interpolation and
truncation.  Those of us who grew up with FORTRAN (which generally made
extensive use of floating-point numbers) are likely all too familiar with
that issue.

Regards,

Glenn

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *