Hi @Jenna_Shelton! You identified a bug in qiime demux summarize where sample IDs with underscores are truncated (and one of the corresponding sequence files is arbitrarily used as the truncated ID's data). I've created a bug report so we can get this fixed in an upcoming release.
For now, I recommend renaming the sample IDs to not include underscores and re-importing that data. In order to avoid the bug you'll need to rename both the sample IDs in the fastq manifest as well as the .fastq filenames on disk. Sorry for the inconvenience!