The optical sensor output is not a true TTL output as is claimed, it appears to be an over-driven analogue signal - in that is appears to be linear over a small range. You might be able to see this with a multimeter measuring the output voltage. Also, the operation of the green LED on the sensor isn’t a reliable indication that the emonTx has or hasn’t detected a pulse, due to that pseudo-analogue output again.
One thing I do notice is, the single-phase emonTx sketch detects a pulse on the falling edge of the pulse (i.e. going dark) and the 3-phase sketch on a rising edge (going light). You can change this on line 394, if you wish, though as the trigger input is level-sensitive, as distinct from slope-sensitive, it should make no difference.
Unfortunately, we don’t have any more information about the optical sensor other than that published here and in ‘Resources’, and there’s no means of adjusting the sensitivity.
The only suggestion I have is to make sure the hole in the middle of the sticky pad is accurately lined up on the point where the LED is brightest, and the sensor is lined up accurately on the pad.