We present putative global minimum energy structures for nanoscopic transition metal clusters, with sizes ranging from N=3 to 100 atoms, described by the original embedded atom potential of Finnis and Sinclair (FS), using their parameter sets for molybdenum and iron, and compare selected results with predictions from semi-empirical molecular orbital (SE-MO) theory via further optimization using the AM1* and PM6 Hamiltonians. We find that, for Fe clusters, the global minima found for FS potential consist mainly of polyicosahedral structures with magic numbers N = 13, 19, 23, 26, 29, 39, 60 and 78, whereas, for Mo clusters with sizes N>30, they are more likely to be bcc terminated by {110} and {100}-type surface facets. We find that the global minimum energy structures obtained for the FS potential are, in general, very good starting points for further SE-MO optimization, although the relative ordering of the resulting structures by energy compared to those obtained from global minima of other potentials used to model metal clusters does not in general agree.