3 points is really not enough. The problem is the corner of the bed you did not probe is going to magnify any error in the three points you did measure. Probing 3 x 3 works well, but I do get slightly better results doing a 4 x 4 matrix. I think part of the reason is any error gets more of a chance to be averaged out.

I would suggest printing one of those calibration prints that puts a few large squares down close to the edges of your print bed. You will be able to see if the amount of measurement error you have makes a difference between doing a 3 x 3 or a 4 x 4 probe matrix. If the first layers go down fine either way, it may not be worthwhile to do a 4 x 4 probe on most stuff.