To begin with, you do have sag. Is this affecting operation at all right now? Do you see an effect other than on the meters?
It's likely to be bad long term, especially at higher power, because it is similar to running the internal inverter on single phase, which it may not be at all rated for.
Dave is correct, there are unknowns, and a wiring diagram might help. However, an RPC is generally able to cope with unknowns pretty well, it is not a complex device, and the few factors affecting voltage are reasonably simple as well. There are not very many ways to set one up, and once it is working with reasonable output voltages, it is fairly certain that it is set up basically correctly.
That said, the load you have is not like a motor. All the Fadal info I found shows an inverter between the motor and the supply line (typical for CNC), so it is a rectified load, different from a motor load. There are other loads as well.
What works for a motor load may not work as well for an inverter load, because a rectified load is hard to drive. The rectified load has pulses of current considerably higher than the average, with a time between pulses, due to the rectification. The pulses can be from 2 to 5 (or more) times higher than the average or rms current, depending on source impedance and other factors in the supply itself.
That can cause an effective voltage drop for any supply phase with a higher impedance.
If the unit has a power factor corrector (PFC) on the input, which it may, then the load will be considerably "nicer", less hard on the supply. I don't know if that unit has a PFC input. The PFC will eliminate the high current pulses, instead drawing current much the way a plain resistor does. It "shaves off" the high peaks of current characteristic of a plain rectifier input.
Your meters are clearly measuring delta voltage, not "wye". Two remain pretty constant, while one sags quite a bit. That's not necessarily typical, as the voltages are all related. It's more typical to find one steady, and two that sag, but phase shifting due to the load can change that.
It isn't clear ho the current is being measured, but the easiest way is just in the supply lines, so that "wye" current is measured.
I would assume that the top voltage is the input, measured between the single phase wires, and the lower two are from input wires to the manufactured phase wire. This is because the top voltage stays steady under load, which a low impedance source would do.
It may be that a 'balance" capacitor change can help, and it's cheap, so worth a try.
It may also be the case that the load is obnoxious enough that the 10HP idler has too much impedance. With RPCs, it is often the case that a bigger idler cures all problems, because the problems are due to idler impedance, and a bigger one has less impedance. A bigger RPC acts more like a power company 3 phase supply
For a bigger idler, your medium load might be just a very light load, and the heaviest load might be less of an issue than the medium load is now.
Right now, you are getting quite a bit of sag (about 10%) on what is really only a couple HP load. I don't know exactly what would happen with a load much closer to 5 HP, but I very much doubt it would be any better.
Just throwing numbers around, assuming that the inverter load draws current pulses about 3 to 4 x what a motor would draw, one would come up with the idea of using an RPC of about 3 to 4 x the nominal draw. In this case a 20 HP idler.
I would NOT run out and replace what you have just yet, but keep that in mind as you try the balance caps.