An ultrashort composite cavity (</=10 mum in length) laser diode was used to verify experimentally a butt-coupling model that considers interference inside the external cavity and its influence on laser operation. The fine structure of an experimental coupled power versus laser-to-fiber separation curve is explained when a complex coupling coefficient is introduced into the model. The importance of the phase term of this coefficient and regions of validity of the modified butt-coupling model are also discussed.