My comments:
1. Anyone who is making a decision among Macsyma, MathCad, Derive
based on which is fastest on factoring integers should re-examine
his/her assumptions.
a. If factoring integers is really important, there are much
better (free?) standalone programs.
b. If factoring is not important, why use it as a measurement?
It is certainly NOT the case that a fast integer factoring algorithm
means that other algorithms are proportionally faster.
2. Factoring algorithms written in C or assembler can ordinarily be
hooked into a lisp system, if for some reason one needs to both
factor integers and ALSO do (say) polynomial arithmetic. As an
example, I am using a lisp that I read GMP into. (GCL, sometime,
may have GMP in it. Mathematica has GMP in it now. does CLISP use GMP?)
Then see..
http://www.frenchfries.net/paul/factoring/source.html
I suspect that the original Macsyma system had a factoring algorithm
partly written
in assembler for the PDP-6 computer.
3. Systems that have the rudiments of a CAS and also do factoring
are available (free) like NTL. They might be of interest too.
4. If Maxima was slower than Derive, your test might run differently
if you used a different version of Maxima, namely compiled for a
different Lisp. CMU-CL vs GCL vs CLISP vs Allegro vs SBCL ...
And the performance on numerical code varies substantially.