A few things to keep in mind:
Ground is not special. Not in reality, and not in LTSpice. Ground is nothing more than the potential that we've decided to be 0V. It's a label, and one that is totally contrived and arbitrary.
To drive my point home, it doesn't matter what part of your LTSpice circuit you pick as ground. If you move your ground from one net to a completely different one, there will not be any change in the simulated result. The values will probably change but superficially only (because you've changed what LTSpice is using for 0V).
LTSpice can only simulate one circuit. Isolation or floating nodes are not supported.
That said, it sounds like you might be overthinking this. The only thing that you need to worry about when choosing your ground node is what you want LTSpice to reference all the voltages in the simulation to. That's all.
And when you want a 'second ground', what does that actually mean? It means you simply want a net that is, for all intents and purposes, not connected to ground, but is kept at the same potential. 'Kept at the same potential' here really just means that you want this to also be a 0V reference point.
What I typically do is use the already available 'COM' net option, which is just another net label and symbol provided for convenience. It isn't connected to ground, its just connected to what you connect it to. I build my circuit exactly how I intend it, with the separate GND ground and COM grounds placed and connected just like they would be physically.
Then, once I am done, I connect COM to GND... though my trusty 1 EΩ resistor. That's right, Exaohms. Is that perfectly isolated? No, but neither is your real-world circuit. The leakage through our 1EΩ resistor is going to less than an fA, which is likely substantially (like, orders of magnitude) less than the leakage you'll be getting in the real deal.
But don't just use a resistor, put a 1 zF (yep, zeptofarad) capacitor in parallel. This will again be much much lower than the real capacitive coupling that is almost certainly present when this is built physically, and it eliminates some issues with unrealistically high resistance values making simulation speed extremely slow.
Of course, in your application, it would probably be better try and make a rough estimate of the parasitic capacitive coupling you might have between your power ground and chassis ground and use that value instead of a 1 zF capacitor. a few pF is not unusual.
Here is an example of this in action. It's the text fixture for an isolated push-pull power supply. Note that the isolation is simulated using COM on the output, but with this little impedance hack, it still behaves exactly as expected.
Regardless, it really is that simple. But it is also easy to convince ourselves that it isn't.