Manuel pixel level adjustments have always been difficult, even with a mouse. Place elements roughly by eye, and then have nudge buttons to position exactly. Have good support for snapping and alignment in a UI that people actually understand.
Four buttons that nudge an element by one pixel is perfectly accurate. For most of us aligning things on screen is not an exact science. Drawing is a process of minor adjustments, not a single operation. There is scope for software trying to guess the intent, and ignoring the exact placement. This is difficult as snapping can be very annoying when it is not done right.
