As far as I can tell, LT Spice does everything right.
One can use The Thompson Trick to demonstrate that LT Spice is wrong: run a sim, change the schematic but don't rerun the sim, screen capture, complain that the screen shows a conflict. You can do a power calculation after changing the schematic and the RAW file data doesn't change, so you can manufacture criticism that way too.
If you change the schematic, it's a good idea to rerun the simulation. I suppose LT Spice could delete the plot plane and delete the RAW file if one makes any change to the schematic pane, but I prefer it the way it is. I'm not working to deliberately demonstrate that LT Spice is wrong; you are.
Can you post an .asc file that simulates negative power in a resistor, without using The Thompson Trick?
Incidentally, resistors are indeed "signed" and operate correctly, and the generated power equation always seems to have the correct sign. To tell which direction a resistor reports current, run a sim and then hover over the resistor. The cursor becomes an arrow that shows the current sign convention.
More than half the time, when I put a new resistor, it's in the wrong direction by my standards, so I flip it to make the current direction to my liking. If you create a new resistor the convention is "down" and then if you ctrl/r rotate CCW, the current convention becomes left to right, whereas I usually want right to left; so I ctrl/r three times.
Always glad to try to help an amateur get un-confused.