Context. In a very recent measurement Gurell et al. (2009, A&A, 508, 525) commented that while the theoretical lifetime of a 4 G 5.5 is approximately one tenth of the lifetime of b 2 H 5.5 the corresponding measurement shows this to be close to one fifth. This large discrepancy is attributed to the effect of inadequate level mixing in the theoretical calculations. Aims. The aim of this work is to make a detailed analysis of these level mixing effects on transitions from various lower levels to the a 4 G 5.5 and b 2 H 5.5 levels given in three previous calculations and in the present more extensive CI calculation. Methods. The CIV3 structure codes of Hibbert (1975, Comput. Phys. Commun., 9, 141) and Hibbert et al. (1991, Comput. Phys. Commun., 64, 455) are used in the present work, combined with our "fine-tuning" extrapolation process.Results. The calculated mixing between the upper levels, obtained in previous calculations, is shown to be too weak. The stronger mixing determined in our work gives rise to a calculated lifetime for b 2 H 5.5 within 3% of the measured value. On the other hand our calculated lifetime for a 4 G 5.5 is around 20% lower than the measured value, which has fairly wide error bars. Conclusions. Our enhanced calculations explain the difference between previous calculations of the b 2 H 5.5 lifetime and the recent measured value and confirm the latter. We also suggest a somewhat higher value than experiment for the lifetime of a 4 G 5.5 .