I would suggest putting a 1×Salmple loading buffer in any wells that you did not add samples. This will make the samples run more uniformly. If you can load less sample or use a thicker gel 1.5 mm instead of 1 mm. Test your sample buffer by mixing it with a protein ladder and running the gel. If the same problem arises, change the loading buffer. Electrophoresis conditions seem fine, I use 25 mA per gel, which gives me ~90 V for stacking, and increases over the course of the electrophoresis.
This occurs either because of a high sample load leading to gradual sample spreading to the side of the gel while performing electrophoresis or too harsh an assembly of the transfer, leading to deformation of the gel. I would recommend loading the same volume in each well (ag 15 µL) by diluting the marker and samples loaded less than the designated volume with 1×loading buffer. In addition, load 1×loading buffer in any empty wells.