There are a couple of limitations to how we capture visual rendering time as the analysis we do is based on the final frame of the fimlstrip. Here's a few reasons you might see rendering time slower than you expected:
One limitation of hero rendering times is that they are affected by elements over the top of an image. When doing the visual analysis we use the last frame of the filmstrip as the final state of the page, we then crop down to the area of the image or H1 and use that to search across the filmstrip and find which time first matches that final state of the page.
This means that when elements are layered over the top of each other the rendering time we capture is for the combination of visual elements.
When looking for a visual match for rendering times we also take into account its position on screen so if an element moves around the screen that can delay the time we report.
Elements that move around in the page as it renders can be distracting for users so may want to look at stopping the H1 moving around which would also improve the rendering time we capture.