@Amanda_Berman, it looks like survey.csv and choices.csv are encoded as Unicode (specifically UTF-8).
Stata 14+ supports Unicode. However, odkmeta was written to be compatible with Stata 11, and it does not currently support Unicode. We may add Unicode support in the future, but for now, I recommend exporting your CSV files as ASCII instead.
From a first glance, it looks like the one character in your XLSForm to watch out for is ’, the right single quote, which doesn't appear in all eight-bit extended ASCII encodings. As long as that character only appears in labels, the simplest solution is probably just to replace it with the straight single quote ' before exporting. The presence of the right single quote may be why the CSV was exported as Unicode in the first place.
I'll create a GitHub issue noting that we should consider adding Unicode support to odkmeta or at least better documentation.