The issue with the z axis is that normally your GDOP (Geometric dilution of precision) is terrible in that direction. GDOP (or HDOP/VDOP (horizontal/vertical DOP)) is a term commonly used in GPS, it’s a measure of how the locations of the satellites (or in this case anchors) is impacting your accuracy.
The maths to calculate the numbers involves a few matrix calculations but the basic concept is simple. Lots of anchors evenly spread in all directions gives a good DOP, all of them the same distance in the same direction gives a very poor DOP.
If you put a set of anchors on the ceiling and walk around underneath them then the maths works out that your height accuracy is going to be terrible. Nothing you can do about it.
Varying the heights of the anchors helps but realistically you’re never going to get z anywhere near as good as x and y, it’s easy to get 20 m of horizontal variation but very tricky to get 20 m of height variation.
One other approach I’ve seen to an accurate Z value is to put a barometer on the rover and the anchors. Barometers are terrible for absolute height but good for relative height, if you know the height of the anchors and the air pressures there and you know the air pressure at the rover then you can make a fairly good estimate of height at the rover.