I just realized you were probably referring to the example given from the AnandTech article with `lea r64, [r64+imm8]`.

Caveat is just that [presumably] the source and destination registers have to be matching (since `lea rax, [rax+imm]` is just `add rax, imm`).