Hi,

There may not be a fixed value for similarity threshold in molecular phylogenetics analysis. If there are similar reports (on ITS analysis) of these or related genus, you  may use similar values for species differentiation. To be more complex (and actually more appropriate), you study or analyse (bioinformatically) the ITS sequences from several related genera and try to estimate the site divergence or rate of divergence in these sequences across related species. You may utilize that information for your problem too. A  researcher more thorough in molecular phylogenetics may be of more help here.

For finding the closest sequence, you don't need any threshold, and no, there is not this possibility in MEGA. It should be noted that closest in term of maximum identity and closest evolutionary speaking may be different as in evolution different base changes happens at different rates; this is taken into account in phylogenetics methods (especially ML), while the percentage of identity does not account for these differences. So I would suggest to collect the most similar sequences through a blast search against an appropriate database, and then building a phylogenetic tree using some ML method. MEGA can be a good start.