If it takes 4 equations to find x,y,z and t, how does one improve the accuracy by adding 5th equation or does it reuse any 3 previous equations and recalculate with the 5th equation?
The quick answer to this is: least squares estimation
Least squares will use all possible observations to produce an optimum position solution from all the observed satellites. Put briefly - the greater number of satellite observations gives greater degrees of freedom with greater redundancy thus reducing uncertainty of position, and furthermore greater likelihood that the observed satellite geometry is stronger.
For a more mathematical description of least squares GNSS estimation see this paper by Geoff Blewitt: Basics of the GPS Technique: Observation Equations