Despite recent progress in the growth of InN-rich InxGa1-xN alloys, the composition dependence of the InGaN bandgap and the size of the InN gap remain uncertain. We apply a combination of techniques, Electron Probe Microanalysis (X-ray fluorescence spectroscopy (XRF) in wavelength-dispersive mode) and Cathodoluminescence (CL) spectroscopy, to the first of these problems. Our method measures in situ the composition and the luminescence spectrum of almost coincident volumes of sample, of size about one cubic micron. The combination of microcomposition mapping with CL spectrum imaging produces very large E-P(x) datasets. (E-P is the peak energy of the emission band.) We discover an unexplained systematic difference in the E-P(x) dependences of samples grown by Molecular Beam Epitaxy and Metalorganic Vapour Phase Epitaxy with similar ranges of x from near zero to similar to0.4. The linear relationship previously established between the bandgap energy E-B, measured by absorption spectroscopy, and E-P for MOVPE samples allows an extrapolation of the MOVPE E-B(x) data to x = 1, representing pure InN, which yields a predicted gap of 0.7(1) eV. This is likely to underestimate the true value. (C) 2004 Elsevier B.V. All rights reserved.