You have to differentiate between corrections that are sources of error and corrections that are well understood and can be accounted for.
Time dilation is a well understood effect, and if you wanted to do something like sync two clocks on earth by having a satellite simultaneously send signals to both clocks (a pretty common timing-metrology technique), you definitely need to account for the time dilation between the two earth clocks and the satellite. And you can do that by tracking the satellite's position and using the relevant equations to find the expected time dilation effect. We aren't in the business of spending several million to put a satellite into orbit, but putting it at 3200km because "ehh... I don't want to write the code to account for time dilation."
However this can become a source of error if there is any uncertainty on the position of the satellite. In which case, actually, you'd want your satellite to sit in a place where the slope of this curve is smallest - not where the total effect is smallest. Since the horizontal axis is logarithmic, it's always better to be higher.
To prove this, say we had a measurement of the satellite's height $h_0$, and some error on that height $\delta_h$. And let's say the time dilation at a height $h$ is $T(h)$. Then the error on the time dilation effect is $T'(h_0)\delta_h$.
All that being said - I do agree it's interesting, and surprising to me, that there's a height at which circularly orbiting satellites have zero time dilation relative to Earth's surface.