The BNL-AGS E906 experiment was the first fully electronic experiment to produce and study double hypernuclei with large statistics. Two dominant structures were observed in the correlated π − -π − momentum matrix at (p π−H ,p π−L ) = (133,114) MeV/c and at (114,104) MeV/c. In this work we argue that the interpretation of the structure at (133,114) MeV/c in terms of 3 Λ H+ 4 Λ H pairs is questionable. We show, that neither a scenario where these singe-Λ hypernuclei are produced after capture of a stopped Ξ − by a 9 Be nucleus nor interactions of energetic Ξ − with 9 Be nuclei in the target material can produce a sufficient amount of such twin pairs. We have therefore explored the conjecture of Avraham Gal that decays of the 4 ΛΛ n may be responsible for the observed structure. Indeed, the inclusion of 4 ΛΛ n with a two-body π − branching ratio of 50% in a statistical multifragmentation model allows to describe the E906 data remarkably well. On the other hand, a bound 3 Λ n nucleus would cause a striking structure in the momentum correlation matrix which is clearly inconsistent with the observation of E906.