Looking at the screengrab, the ‘bad’ caption seems to be left padded with space compared to the two captions beneath it. I’d expect all three to left align (unless there is centring/justification in play). So, is it a leading space? A tab? I wouldn’t be surprised if that is part of the issue. If you edit the source to make sure nothing is in front of the first letter of Restitution, does the OPML then import OK into Tinderbox. If doing this do keep the original Mindmanager file before editing the caption so you’re preserved the original state for later reference.
If so, it is a matter of working out what that character is. If it is is a space or tab then, for OPML/XML, Mindmanager is mis-encoding them. That’s not to point a finger but to work out what need’s fixing. Regardless of any potential changes in Tinderbox, Mindmanager would do better not to include (what appears to be garbage).
It is quite possible the eco-system of mind-mappers get around the problem by ignoring it. Still, if fixed at source, no kludge fix is needed downstream. It would be interesting to know what the Mindmanager dev’s take is here.
The cruel irony in this inter-app context is apps that best-guess so as to achieve a result appear more competent because we see an outcome rather than an error. Of course, especially in a large doc, we’ve no idea what is being disregarded or deleted in order to ‘just’ work. IOW, fist impressions can be deceiving. In saying that, I’m not disparaging anyone here, just trying to drill down and figure what needs fixing—which might not be what or where we first imagine!