I think it may be important to take note of the front end in use. For
some situations, the
length of the transcript of the session so far can affect the time. I
hope this has not
affected the current question/answer thread.
\begin{speculation}
It is possible that several lisps have the same style of garbage
collector and
that the particular operations being timed play into a good or bad situation
for that style. For example, the 'conservative' garbage collector may
suffer
sometimes if memory-in-use gets scattered, since it does not compact
storage. Generation scavenging / compacting garbage collectors
are often much superior though they too can be put in a disadvantageous
situation.
Since many people don't run Lisp systems for very very long times, the
differences don't show up, and so the performance seems the same
in benchmarks, even though one system may be "asymptotically"
quite different from another.
If someone could identify the styles of GC for each of the tested
lisps, and also make sure that they are running the same exact
version of Maxima, that might help.
\end{speculation}
RJF