Other folks should be following this thread too so they will jump in.
Nothing will make sense or be relevent when a battery cable is disconnected. You didn't tell me which cable you measured the 3.5 volts on. If you put one voltmeter probe on the negative battery post, and the other probe on the engine block when a helper tries to crank the engine, you absolutely must not read more than 0.4 volts. If thats the circuit where you had 3.5 volts, there is a loose, dirty, or corroded connection someplace. Moving the probes toward each other one mechanical connection at a time will identify the bad connection.
If I'm misunderstanding where you're finding that 3.5 volts, and you actually have less than 0.4 volts drop on the negative side, do the same thing on the positive side. Put one probe on the positive battery post and the other one on the large stud on the starter, not the nut or cable terminal. You might need a second person or a clip lead unless your arms are real long.
Just to verify, the starter relay is still rapidly clicking, and battery voltage during cranking is still over 10.0 volts. If this is wrong, tell me what's happening.
One thing that might be having an affect on finding this problem is taking voltage drop measurements when the starter relay is chattering. The high current to the starter is turning on and off so the voltage drop will be too. That can make the voltage reading bounce all over and be confusing. To avoid this, you can pop the starter relay's cover off, then squeeze the contact with your finger. That will keep the starter engaged so your voltmeter readings will be accurate.
Also double-check what happens to the headlights. They will flicker but should stay fairly bright. If you find they go out or very dim, that suggests the starter is shorted and drawing unusually heavy current. That will cause the starter relay to chatter due to the battery voltage being drawn down, but that would contradict the 10.4 volts you found earlier.
Caradiodoc
Tuesday, March 30th, 2010 AT 2:34 AM