For x∈[0,1] we have x∈[0,1], so θ∈[0,4π] and therefore 2θ∈[0,2π]. Keep that interval in view — it is what makes the last step legal.
Transforming the argument of the cosine inverse.
1+x1−x=1+tan2θ1−tan2θ=cos2θ,
which is the standard double-angle form. Hence
cos−1(1+x1−x)=cos−1(cos2θ).
Unwrapping. The principal value branch of cos−1 is [0,π], and we established 2θ∈[0,2π]⊂[0,π]. So cos−1(cos2θ)=2θ exactly, with no fold-back correction, and
21cos−1(1+x1−x)=θ=tan−1x.
A numerical confirmation, and a warning about the square root. At x=41: the left side is tan−121=0.46365, and the right side is 21cos−15/43/4=21cos−10.6=21(0.92730)=0.46365. They agree.
The radical is not decoration. If the left side were tan−1x instead of tan−1x, then at x=41 it would be tan−10.25=0.24498, nowhere near 0.46365 — the statement would simply be false. The radical on the left-hand side is essential, not decorative.