Trees constructed by treeing algorithms often suggest a precision which does in fact not exist in reality. To make it impossible to draw wrong conclusions from your tree, you might consider to multifurcate specific branches, especially such with low bootstrap values.
In the 'Multifurcate tree' window you can restrict which branches shall be multifurcated by specifying lower limits for
bootstrap and/or

branchlength.
All branches below the specified and selected limits will be multifurcated, i.e. their length will be set to zero and their bootstrap value (if any) will be removed by clicking on the 'Multifurcate' button.
This is only applied to terminal (i.e. leaf) branches, if 'Also apply to terminal branches' is checked.
The algorithm does not change the sum of the branchlengths of the tree (which is a measure for the phylogenetic distance between all tree members). To achieve that the algorithm distributes the former branchlength of each removed branch to its directly adjacent neighbour branches.
Removing the branch C from this example tree
B
+
A 
+ D
 +
+
C +
E
will result in a multifurcated tree
b
+
a 
+ d
+

+
e
with modified branchlengths, where
SUM(a,b,d,e) = SUM(A,B,C,D,E)
The distribution is performed proportionally to the length of the target branches of the distribution, i.e. longer branches receive more of the distributed length than shorter branches.
Branches with zero branchlength are never changed, as they already represent (and will remain) multifurcations. Instead the length is distributed to the next adjacent branches beyond the multifurcation. This is true for existing and for just created multifurcations.
The distribution simultaneously takes place at all affected branches, always considering the former branchlengths for calculating the proportion and avoiding summarizing effects at repeatedly affected branches (i.e. at branches with multiple removed neighbour branches)
