Best distance of anchors to metal walls

When mounting anchors on walls of a corridor-shape area (metal walls on both sides of a corridor) we experienced a weird problem:
For distance 30cm anchors from wall, we sometimes read huge biased range data (up to 1m bias error in TWR data)
when we mount the anchor close to the metal wall, bias error vanishes but the range data starts fluctuating (up to 15cm fluctuations)

What should be the distance of anchors from reflective surfaces? Any principle behind this behavior?