Wrong Output when Masking in VR

Hi,
I doing a virtual 360 city tour, where I put graffiti on empty walls. It worked fine so far but with a few videos the output is wrong. It always shows up, when i am masking something in the video. For example, if leaves are in front of the graffiti, I am masking them and put them on top of the layer with the graffiti.

My workflow is the following (i work with the VR comp editor):

- Adding 2D Layer

- Removing the tripod

- Adding 3D layer

- Adding the graffiti (png)

- Either add a new layer on top and masking or

- Output and Export to Media Encoder

I uploaded a picture of the wrong view and a picture of the right view.