Just 1nF will not work if you really have 250Hz on top of 50Hz. I would assume that 250Hz amplitude is significant enough to throw off your 0 detection. You could do it in software, but it would require some work to build some virtual PLL. Other option would be active analog filter, but same as with any filter you'd have to be mindful of phase shift it may introduce.
You could investigate why do you have 250Hz interference in the first place and solve problem at it's source, perhaps bad PCB routing?