I can understand the twisted pair idea but a floating ground wire is supposed to help reject noise
The whole idea is wrong. For a start, twisted pair works very well in differential signalling, but that's not how a single ended phono connection works.
Differential signalling is used all over the place from studio audio connections ((Balance Line); HDMI cable; even our wired telephone system and Ethernet network cables use it. In fact that's a good example because telephone cable and Ethernet cable is largely unshielded too. But all of these applications do something important with the signal both before and after transmission down a twisted pair: They all send two versions of the signal, a +1 and a -1, they're 180 degrees out of phase with the other. It's two versions of the same signal and no ground. That's what each pair carry. Summing the two signals together at the destination end without first inverting the phase of one would result in them cancelling each other out. That would leave nothing but the noise acquired on the journey down the wire. Re-inverting the phase of one leg means that it's the noise that is then out of phase on each leg instead. The noise cancels out when summed.
A phono connection doesn't use two live connections. There isn't a +1 and a -1 version of the signal no matter how fancily the wire is wrapped around itself. What comes out of a phono socket is the signal on the centre pin (the '+1' if you will) and a ground. That's it. Any noise that the cable picks up will be transferred to the signal. Plaiting it with the ground or any number of additional dummy floating grounds ain't going to do a damned thing to change that.
The idea of the plaiting isn't anything to do with external noise rejection, so the promotion of it as some magic bullet for noise reduction is total BS. Plaiting helps reduce inductance. Put two wires parallel to each other and in close proximity. Sending a signal down one creates a magnetic flux in the wire. The presence of that magnetic flux induces a current in the second parallel wire. It's the principle upon which motors and generators and transformers and electromagnets are based on. In signal cables we call it crosstalk. Moving the wires further apart reduces the effect. That's all the PVC ICC dielectric does. It's a spacer. Once again though Ixos is barking up the wrong tree. For crosstalk to be a problem there has to be two sets of signal conducting cables in close-enough proximity for the induction to create a ghost of one signal on to the other. That's not happening in a single phono to single phono lead. Any crosstalk from the signal to the ground will get dumped to the chassis earth at the destination-end. Any crosstalk from external noise picked up by the ground wire might well be kept from affecting the signal core, but that's hardly any comfort when the signal core is already exposed to external noise of a greater magnitude.
The version of this cable I linked to does have a shield. It's Mylar foil. Mylar is basically plastic tape covered in a vapour deposited aluminium layer a couple of microns thick. It's the stuff that metallised party ballons are made from. That begs the question "If the noise rejection of this Gamma Geometry cable is so good, then why do some of the cables need an extra shield layer?" I'll let that question hang for a while....
There are two other problems with Mylar. First, this type of shielding only starts to become effective at frequencies well above the audio spectrum. So any induced noise from transformers and leaky circuits is going to pass right through and get straight to the signal-carrying core. The second problem is that as far as conductors go, Mylar is crap. There needs to be a far more effective secondary conductor to act as a drain to ground otherwise Mylar on its own is useless. Cheap aerial cable (RG6) uses Mylar but it is paired with a layer of either aluminium braid or tinned copper braid. This takes any induced charge away from the cable shield to ground.
The whole Ixos cable just stinks of voodoo physics.
