Thank you for the detailed explanation of your question Oliver.
The switching losses are modeled as dirac impulses. In order to have a meaningful interpretation of the switching losses, the impulses should be averaged using the "Periodic Impulse Average" block. Refer to the "Buck Converter with Thermal Model" demo model for an example of how to use the averaging blocks to determine losses and efficiency.
In the context of your graph, the "width" of the impulse is actually a visual artifact of the solver. There is a dirac impulse that shows up at one time step, but at the next solver time step (after the switching transition) the switching losses are zero. With the default scope settings, the PLECS Scope will connect the two points with a line, which is what you're observing.
You can change how the trace is plotted by setting the signal type of the non-averaged switching loss signal to "impulse". This is done by opening the cursors to see the data window, and clicking the small icon to the right of the trace name. Again, you should ultimately use the "Periodic Impulse Average" to interpret your results.