Cutoffs for Proxy Identification: D' and R^2

Question

[dec986]

I'm using a proxy-finding script with plink that reports potential proxies with their D' and R^2:

proxy   D'  R^2
9:12345698:rs12345  1.0 0.000758121
9:12345999:rs87654  1.0 0.039958
9:12346999:rs54321  1.0 0.0399958

cf. https://www.cog-genomics.org/plink/2.0/ld

However, what cutoffs should be used? From what I've read, D' is ideal when it's about 1, and R^2 is good when it's about 1 as well. However, these results above seem to contradict each other.