Observations rather than 'facts', I'd think, and rarely are they 'explained' - just predicted. If a model's prediction fits with repeatable observations, then it gets adopted. Then along come some more observations which it doesn't predict. Have the 'facts' changed? No. Is the model 'wrong'? No, it just doesn't fit everything. We don't have a scientific model that doesn't run out of steam somewhere.
It's a repeatable observation - something different from a fact. And of course, the Earth isn't spherical in shape - the spherical model works well for many things, but not for everything. It's a philosophical point maybe - and not to be used to argue against observable, empirical 'facts' in less abstruse areas of endeavour.