if you have any thing higher than 2.5 volts then i would use a 1 watt 3.3 volt zenner diode in series with the led.
I was almost getting around to suggesting a Zener might be needed, but not knowing how many volts needed to be dropped, i started at a basic rectifier diode (0.6v).
also make sure the filter cap for the insert lighting is up to scratch, you will find if this capacitor is stuffed or weak then you will be getting some A.C. ripple on the column and row transistors causing them to pulse at 50 hertz which will make your leds glow, if this is the case then the more incandecent lamps that are lit then the brighter your leds will glow. it's all food for thought. it might just be that as i said at the start or this reply, the designers may have just wanted to keep the lamp filaments warm. 
Definately worth looking into....from what i keep seeing in these old pins, the designs aren't exactly stable, lots of AC of spikes and buzzing on many of the supply voltages...
when trying to work out what voltage you have do it with a analog meter with your pin in test mode, try the individual lamp test then try row and colum test.
Another good point, as Retropin has seen, the Digital Meter gives crazy reading with so much 'switching' going on...
The old Analog will give a nice average reading.....
At the end of the day, anyone with a Cro will see exactly what volts and switching is occuring .....
On another note...
Are ALL the leds glowing? or just the playfield-insert ones????
I wonder if there is still a large amount of side-light/leakage ( when one light illuminates the insert
next to it ).
Placing some small card-cylinders around the lamp-fittings (undernath the playfield) will prevent inserts lighting up from nearby globes....something that REALLY bugz me hehehehehe.
MM.
