And that's the problem, the fonts used by cairographics on Mac OS X.
I don't know about the 'no longer': this (and a few others including
infinity) have been faulty for some time. I think it relates to how
'fontconfig' (for which a binary library ships with R for Mac OS X)
interacts with Mac fonts, but there are many more interesting things
to do that track down such quirks.
There are some hints about this on ?X11
X11 chooses fonts by matching to a pattern, and it is quite
possible that it will choose a font in the wrong encoding or which
does not contain glyphs for your language (particularly common in
'iso10646-1' fonts).
This does look very like an incorrect encoding for a symbol font.
OTOH, the last display in example(points) seems right, so something
subtle is going on.
One option is to use X11(type="Xlib"). Another is to use quartz(),
assuming you are not running R on a remote system.
This is an Mac-specific question and it is always worth asking such
on R-sig-mac.
I believe the behaviour is still the same in 2.9.0 patched, but it
is always worth checking the current release, especially on
platforms where components that are otherwise part of the OS ship
with R (as no one records in the R docs things fixed/broken by
updates to such components -- we probably don't even know half of
them).