It looks like either position of tower X needs to be moved counter clockwise a bit or position of tower Y needs to be moved nearer to the centre. Hard to tell your description is too vague.
But the good thing is
I do not need to guide you since you can read what to do in this document: [
github.com]
Read the comments in it. If you are good at math you can even try to execute it.
That guide (which recommends (current steps * expected size) / (actual size) = new steps per mm) is misleading. It works like htat in cartesian printers. In delta, that works only approximately and only for Z coordinate.
The other guide (which recommends adjusting the diagonal_rod_length in order to get an XY accuracy) is misleading too. That advice may work in some special cases (e.g. when only diagonal rod is wrong and when you print the object at specific location with a specific orientation) but in general it is wrong.
In a linear delta machine, X, Y, and Z errors are all tied together. If your steps per mm are correct, bed is flat, towers are equidistant to each other, and the bed is perpendicular (or almost perpendicular) to the towers then if you perfectly level the bed then also your X/Y dimensions will be perfect (or almost perfect).