Okay, so your concern is low frequency EMF, generated by the current flowing in the conductors. This is about the magnetic field produced by 60Hz current flow, not high frequency radiated EMI.
I'm rambling with theory here, don't expect practical solutions, unless something that I say triggers a practical thought on your part:
The only thing that gets rid of the EMF is having balance current returning in the opposite direction. The closer the phases are together, the better the balance, and the less the EMF.
In a balanced circuit situation, the net current flow of the conductor bundle is zero. There is no net flux coming out of the bundle. There is some flux produced _between_ the conductors. A steel conduit will actually _increase_ this flux, because it is lowering the reluctance of the magnetic path around the individual conductors; we see this in the difference in reactance of circuits in steel rather than non-magnetic conduit. However the steel conduit is also acting to 'short circuit' any flux which escapes the bundle of conductors, so we don't see magnetic flux on the outside of the pipe.
When you have isolated phases, you don't have balanced current in your pipes, the steel conduit acts as the core of a transformer, increasing the flux produced by the individual conductors. The steel will neither increase no decrease the flux seen on the outside of the conduit, with one exception: the increased flux around each conductor will mean increased induced voltage in the region of the conductor, and thus current flowing through the steel itself. This current flow in the conduit will act to balance the current flow in the phase conductors, thus reducing the net current and thus the net magnetic flux.
If you use aluminium conduit for isolated phases, you will have no shielding effect from the conduit, and no change in flux produced. Again, any current induced to flow in the aluminium conduit will tend to reduce the net current and the net flux.
In fact, if the conduit were a 'superconductor', then the current flowing in the conduit itself, induced by the magnetic field produced by the isolated phases, would entirely eliminate any external magnetic fields.
If you provide conductors of sufficient ampacity in your steel conduit, arranged as 'shorted turns' between your various parasitic transformers, then the current flow in these shorted turns will act to reduce the net flux. Lots more copper than the original installation, but you could use steel and isolated phases, and not have much external flux
Steel that is not _between_ the conductors will not be subject to significant inductive heating. The worry is large current loops parallel to the phase conductors. I do not believe that steel that surrounds the _entire_ set of phase conductors will be an issue. I don't think that you will have a problem with rebar that wraps around the duct, though rebar that runs in the direction of the individual conduits and next to the conduit may be a problem.
I suppose that you could wrap the entire duct bank in a steel shell, enclosing all of the conductors and the current loops. This would provide shielding for the EMF and be outside of the region where you would have significant inductive heating. This shell need not be structural, and possibly not even continuous; dense re-bar on the outside of the duct, combined with non-metallic conduit for the conductors of the duct itself might provide the necessary shielding, not have inductive heating, and permit isolated phases.
-Jon
