A few comments....
- it looks like this is posted twice....the moderators will not be happy with double posting
- Its not clear the magnitude of RPM you are measuring.
- Is it worth using microseconds vs millis for better accuracy
- It may be worth disabling interrupts for longer, in case you are getting multiple interrupts at the same time (de-bounce)
- you are likely to get some interference from ambient light if the sensor is not physically shielded
- probab;y a good idea to mention the model numbers of the LEDs
- It appears your input is analog and you are feeding this into a digital interrupt pin....not surprising you would get varying results?
etc.....