That's not necessarily a problem. Keypress Monitor doesn't send the keypresses through the debounce algorithm, so some ghosting can happen in keypress monitor that is not gonna show up in the real keyboard. In particular I noticed that non-pressed keys can light up that are in the same logical row as pressed keys. (see the matrix layout to look at the keymatrix in its logical form). This seems to be a property of capacitive sensing. A couple months back I added some anti-interference code that reduced by about 50% the occurrance of such interference even when not using debouncing, but looks like it's still possible. I'm still thinking of ways I might improve it further. In any case, it shouldn't be necessary, since debouncing should take care of that. I'm also thinking of just adding some form of debouncing to Keypress Monitor, to have it enabled by default, to avoid scaring people.
To continue debugging, please PM me the following:
1) A screenshot of the Signal Level Monitor after it has been open for a couple tens of seconds with no keys pressed.
2) A screenshot of the Signal Level Monitor after all keys have been pressed slowly (you can press more than one key at once, just pay attention to not go too faste due to the slow refresh rate)
3) A screenshot of keypress monitor, mostly to see the trained thresholds.