Yes, a similar topic reported this behavior the other day. There seems to be a possible bug in emperor that causes it to fail to display plots containing only 2 samples, and it looks like @yoshiki is on the case. In the meantime, the workaround is to include > 2 samples...
Welcome to the forum @fulbag! Yes, the reads being different lengths and from different V regions will cause issues in downstream analysis. @cjone228 and others have been discussing strategies for processing these data throughout this topic.
Splitting up the data by V region and processing separately seems to be the consensus suggestion that I and others have recommended in the past. Other methods, like using SMURF, sound like they have been troublesome and are not currently possible to adapt to other protocols.
Another possibility would be to use closed-reference OTU clustering to essentially merge all V regions into full-length reference 16S sequences. You may lose some of your sequences (if they poorly match the reference) but it is worth giving it a try...
Disclaimer: I do not have personal experience processing ion torrent data, and @cjone228 and others may be able to offer more guidance... I just want to make sure your question is not lost in the fray!