You can (at least in theory) find the position of a star in a photograph of the sky to an accuracy of less that one pixel. It's straightforward to do:
-- If there's one bright pixel, then the star is centered in that pixel.
-- If there's two bright pixels of equal brightness, then the star is halfway inbetween.
-- If there's two bright pixels, but one is slightly brighter, then obviously it's closer to that pixel.
In practice, you just find the "centre of light" (by analogy to the centre of mass) of the image. A weighted average of the coordinates of the pixels in the image, weighted by the brightness of each pixel.
Presumably your device is doing a similar thing, except it's even simpler because it's only 1 dimensional.