My professor gave me the following derivation for the full generator of the Lorentz transformations. The starting point is to consider a subgroup of the conformal group that leaves the origin fixed (as done so in Di Francesco, for example). Denote by $J_{\mu \nu}$ the full generator and so $J_{\mu \nu}\Phi(0) = S_{\mu \nu}\Phi(0)$, where $S_{\mu \nu}$ is the component of the full generator transforming only the field spin components.
Infinitesimally, $\Phi'(0) = (1 + \epsilon S_{\mu \nu})\Phi(0)$ but also $\Phi'(0) = e^{-ix' \cdot P}\Phi'(x')$. This I understand. The next line is the first one I am not so sure of. He then writes $$\Phi'(0) = e^{-ix' \cdot P}(1+\epsilon J_{\mu \nu})\Phi(x)$$ How is this true? He seems to be using the transformation $\Phi'(x') = D \Phi(x)\,\,\,(1)$, but using $D$ as the full generator. If I had read that statement a few months back, I would have accepted it fine, but I posted a question here earlier that discussed the fact that in $(1)$, $D$ is the spin component of the generator because $x'$ and $x$ represent the same physical point in the space, they are just two different representations of the same point relative to two different coordinate systems. So the statement seems to be contradicting what I thought I understood. That was my first concern.
My second concern is less fundamental (more of a computational problem) and I can edit it into my answer later. Many thanks!
Edit: So assuming the last line to be correct, we proceed: $$\Phi'(0) = e^{-ix' \cdot P}(1 + \epsilon J_{\mu \nu})e^{i x \cdot P}\Phi(0) \approx (1 + \epsilon \underbrace{e^{-ix \cdot P}J_{\mu \nu}e^{i x \cdot P}})\Phi(0)$$ using the fact to zeroth order $x' \approx x$. The brace can be evaluated using the Hausdorff formula: \begin{align}e^{-ix \cdot P}J_{\mu \nu}e^{i x \cdot P} &\approx J_{\mu \nu} + [J_{\mu \nu}, ix^{\sigma}P_{\sigma}] \\&= J_{\mu \nu} + i\left[x^{\sigma}[J_{\mu \nu},P_{\sigma}] + [J_{\mu \nu}, x^{\sigma}]P_{\sigma}\right] = J_{\mu \nu} + x_{\mu}P_{\nu} - x_{\nu}P_{\mu},\end{align} where the second commutator on the RHS there vanishes. Now compare with $\Phi'(0) = (1+\epsilon S_{\mu \nu})\Phi(0)$ as written at the top and we have the result $J = L + S$. My question is: why does this second commutation relation vanish?
I should say I am going through 'Conformal field theory' Di Francesco section 4.2.1. The above argument was an alternate argument my professor posed when we sat down and agreed that equation (4.26) in Di Francesco is simply wrong.