Nice experiment, Jeremy.
From your pictures, your top wouldn't seem to me one with a so low CG to stay leaned always towards the heavy side.
Thanks, Iacopo! I think that for any given top with static unbalance, the phase angle in effect at the time of observation depends on a combination of top properties and speed. The top properties involved include CM height, AMI, TMI, and the magnitude of the unbalance. The fact that topple speed also depends on these properties (especially AMI and CM height) only adds to the confusion.
The LEGO tops I've tested so far show just one phase angle (180°) throughout their operating speed ranges (~6,400 RPM to topple speed). In practice, however, it's impossible to know what phase angle might do below topple speed, and that "blind spot" only grows with CM height. The top shown in my post has a minimum topple speed of ~480 RPM. There could easily be another whirling mode with 0° phase angle below 480 RPM, but I'd never know it.
I suspect that the tip of your top was slipping on the spinning surface, because of high speed and slippery surface, (if you used the glass pane in the picture);
if the tip slips on the base, the direction of leaning becomes unreliable.
That may well be true of some tops, but I haven't observed any variation in phase angle due to tip and surface properties alone. For any given top, I get the same phase angles on the glass table shown in the photos, on concave glass lenses of varying curvature, on finished wood, on lids from plastic containers, on coarse- and fine-grained polished granite, and even on polished granite covered with 1.8 mil plastic film, which, due to tip sinkage, tends hold the contact point in place.