On the interpretation of the multicolour disc model for black hole candidates