… Working LED read 1.745 volts.
Right. So if the diode-test current was 1 mA (as Fluke and others use), that's a whopping power of less than 2 milliwatts (1.7V x 1mA). It would take quite the imagination, or stunningly good eye-sight, to see 2 milliwatts even faintly illuminating a 700 milliwatt (14V x 50mA) incandescent filament bulb!
As for the "we_honest" signal, they have been known to use the "common-anode" configuration whereby the 2 "+" wires are tied together. In your earlier photo, it appears Lionel tied the 2 black wires together suggesting a common cathode configuration (assuming black is "-", red is "+"). So you wouldn't be able to use the existing Lionel circuit (presumably common-cathode). In any case, since 3mm red LEDs go for about a penny a piece...and you've come this far...I think it would help the OGR community to go the distance and repair your LED-based crossing signal. In my opinion of course.
Separately, if someone stumbles across this thread in the future, I'd think it would be fairly simple to convert the original incandescent version to the LED version. So if an incandescent version goes out, and you don't want to send the unit back to Lionel as suggested in the manual, consider converting to 1-cent red 3mm LEDs. I'm guessing the incandescent bulb is/was a 1/8" grain of wheat which is effectively the same size as a 3mm LED (1/8" = 3.2mm). Some soldering and such will be involved but I think a 5-cent resistor on the common black wire going to the signal head is all you'd need to add to reduce the ~14V (or whatever your accessory power) down to the ~2V needed by the red LEDs.