Decomposition of the solvable Lie algebras of maximal supergravities in D = 4, 5 and 6 indicates, at least at the geometrical level, the existence of an N = (4, 2) chiral supergravity theory in D = 6 dimensions. This theory, with 24 supercharges, reduces to the known N = 6 supergravity after a toroidal compactification to D = 5 and D = 4. Evidence for this theory was given long ago by B. Julia. We show that this theory suffers from a gravitational anomaly equal to 4/7 of the pure N = (4, 0) supergravity anomaly. However, unlike the latter, the absence of N = (4, 2) matter to cancel the anomaly presumably makes this theory inconsistent. We discuss the obstruction in defining this theory in D = 6, starting from an N = 6 five-dimensional string model in the decompactification limit. The set of massless states necessary for the anomaly cancellation appears in this limit; as a result the N = (4, 2) supergravity in D = 6 is extended to N = (4, 4) maximal supergravity theory.