Hi @dzyla,
The window inner radius and window outer radius in a refinement job control windowing that is applied the the particle images rather than the volume. So in your case setting them to 0.6/0.7 windows the particle images before using them for alignment/reconstruction. The 3D density however is always windowed by a soft spherical mask that just cuts out the corners of the box, along with a shaped mask (by default, based on dynamic masking). In the final refinement output, however, these masks are not applied - you are seeing the full result of reconstruction from the particle images directly. The masks are only intended to have an effect on alignment. So even though you see density outside of the window that you specified, this is normal and not a problem - since that density is not used for alignments.
Hope that helps!