How would you go about constructing the “lines” joining points? Given two points $A=(x_A,y_A)$ and $B=(x_B,y_B)$ you need to find a circle through these two points which has its center on the $x$ axis. You can either go about this by looking at the equation of a circle, or by observing that a circle of this kind will be symmetric wrt. the $x$ axis, so $A'=(x_A,-y_A')$ will be a third point on that circle. Here you have a circle defined by three points, so the line is defined. Right?
No. The above is true in general, but there are several ways how this might fail.
- For example, three points do not define a circle if they are collinear and if your definition of a circle does not include circles wich degenerate to lines.
- Three points also do not define a circle if two of them coincide. The case $A=B$ is usually excluded by the axioms, but $A'=B$ would be a problem for us, too, and this isn't covered. So you can't get a line through a point and its mirror image.
- There is also another critical situation where $A=A'$, i.e. $y_A=0$. You might use $B'=(x_B,-y_B)$ to still define this circle, but if both $A$ and $B$ lie on the $x$ axis you don't have a unique circle either. You can recover from this since the circle with center on the axis through these two points is still unique, but it's a special case worth considering nonetheless.
You might stand a better chance by taking mirror-image pairs of points as the “points” of your incidence geometry. Probably exclude the $x$ axis itself. Then the axioms you quoted should hold. Instead of taking pairs, you might exclude the lower half of the plane altogether; that's the same thing. Actually the geometry you get from this is essentially the Poincaré half-plane model of hyperbolic geometry, so there are several more axioms satisfied by this setup, and it's actually quite useful.