Hi Guilin, thank you so much for your advice and references.
I'm testing the example file from the gallery ("OLED(2D)"), which should be properly settled by default.
Also, when comparing the two methods, I've made sure the materials and thicknesses in FDTD and STACK are exactly the same, which makes the different results puzzling to me.
According to the suggestion, I have tried to accurately place the dipole halfway between two mesh points. Unfortunately, the results are barely affected and are still largely deviated from the analytical solves by STACK.
I would also like to note that the two examples you listed are both 3D-FDTD simulations. The problematic example I mentioned in the post is a 2D-FDTD simulation on dipole emission. I've tested that 3D simulations on LEDs are consistent with STACK results. I find it difficult to find 2D simulation examples on dipole radiation.