I'm seeing the support for the "Earth" tilted at 1:50 and thereafter.
The 5 hour duration is correct:
but obviously it starts when the moon's shadow starts to touch Earth; the experiment begins with the moon already on the Earth, which is wrong for this analysis. To see the full 5 hours, the shadow must transit completely, from start outside to finish outside.
The rationale for starting inside Earth is given at 1:30.
Note that this references the times with respect to a single point of observation, and that the total duration is under 2 hours
and no longer 5 hours. This is the greatest failure.
Also note that the penumbra is already halfway across the globe at this time: people in the central spot on that globe would just see the model eclipse begin at this time. This cut the eclipse path in half.
The modelers are not clear on whether they are observing the core shadow (umbra/total), or the fringe (penumbra/partial); in the scale model, I expect the umbra is much too big compared to reality, and the penumbra is too small (and they're not paying any attention to it anyway). This is a result of them not modeling the sun correctly. If you model the observation from a single point, that is obviously quite sensitive to getting these sizes correct.
They do not show the light source in the video. If the light comes from a near point light source, it is not (near) parallel, as the sun's rays are, and moving the moon with respect to this source will move the moon's shadow more than it would if the light was parallel, coming from a far sun.
Have I overlooked anything?