Antonio,

My intuition about the rod and wire in parallel appears to be incorrect
unless the resistance of the rod is quite low, at least on 7MHz. My
EZNEC model shows that the wires in parallel give substantial loss (up
to a couple of dB or so lost in the resistance of the rod) with the
wire connected at the top and bottom and the total rod resistance in
the 1-10 ohm range.

Separated and insulated from the rod seems to work quite well on 7MHz.
However, if the rod is connected directly to the ground radials a large
amount of loss is introduced. (I got 5dB in some cases with fairly high
resistance)

I have found other frequencies where the loss is more than a dB or so
even in this configuration (18MHz is near a halfwave resonance, which
is problematic, as Reg has correctly surmised).

The model shows that the current distribution on the lossy element is
occasionally complicated and bears further investigation. I've posted
the model I put together online. It's in some random configuration of
frequency, load resistance, and connectedness of the lossy rod to
ground, I don't know where it was when I saved it:

http://www.n3ox.net/projects/eznec/carbonpole.ez

It has too many segments for the free version but it probably bears
more investigation ... it is possible for the setup to be quite lossy
on some frequencies. I think Reg's comment about halfwave resonance
and intermediate resistances between very low and very high was quite
accurate, as was mine about quarterwave resonance and connection to the
groundplane.

73,
Dan