Wrong note values with some dotted notes when beamed (0.9.6 regression)

Tested with r. 2905, but the issue is there since at least beta 2.

Sometimes, when a dotted note is in arsis (spelling?) and beaming is continuous, the note lacks one 'tail'.

The attached .MSCZ illustrates such a case (sorry for the spelling error in the title!).

It contains 2 measures, each contains exactly the same rhythm, the first with default beaming, the second with no beaming. In the first measure, the notes marked in red are drawn wrongly: dotted quavers instead of dotted semi-quavers (try turning their beaming off to check they are indeed defined as dotted semi-quavers).

The attached .PNG is a screen-shot of the same.

Thanks,
M.