We are only interested in one event located at x2, and it is the event represented by (x2, t2). You might as well start representing events with both spacial and temporal coordinates, since you are not making any progress trying to do everything your own way. And we are also only interested in one event located at x1, and that is the event represented by (x1, t1). Of course, by definition, (x1', t1') must be the same event as (x1, t1), where one representation is recorded by system 1, using its own coordinate system and clocks, and the other representation is recorded by system 2, using its own coordinate system and clocks. And because those two systems are in relative motion to each other, it turns out that x1' is not numerically equal to x1 or x2, and it also turns out that t1' is not numerically equal to t1 or t2. What the light postulate really requires is that t1'=x1'*c. If you were correct in your wild assumptions, the LT would have been x1=x1' and t1=t1' and x2=x2' and t2=t2' but they most certainly are not! The class 1 event (x1, t1) occurs in system 2 at (x1', t1') where x1' is not co-located with x2, and therefore t1' is not equal to t2. This does not violate the light postulate provided x1'/t1'=c.