When using a comparator in substraction mode and inputting the same signal strength in both inputs, the result (which obviously might be 0 because n - n = 0) isn't correct all.

Steps to recreate:
1. Place a comparator in substract mode.
2. Place one redstone dust in the back input and one in any of the laterals.
3. Place a redstone torch to connect both inputs in the midle, soe the signal strength is the same in both inputs.
4. Enjoy the results.

If you continue testing different combinations of signal strength you will see how almost any of them are correct either.

I really think this should be a pryority to solve: any complex redstone mechanism doesn't work anymore.