We did this quality control with Ioniser in R, with our paired-end nanopore FastQ data

it's about the Y-axis "mean base quality per read" in the quality graph

What does the "Mean base quality per read mean?? it's to low to be PHRED for instance

You can see it here:

https://gyazo.com/df7cd33fddd1b41a7fab138720fda634

(hope the link works)

This does indeed relate the phred quality score for the base-called data.  

If base calling was to some extent successful each fast5 file contains the called sequence in FASTQ format.  Ideally both strands of a molecule were read and are used together to give a more accurate call.  These reads end up in the '2D' category.  However a fair number of reads don't fall into this category and you have separate information for the 'template' and (potentially) 'complement' strands.  

For every read in each of those bins IONiseR calculates the mean phred score, and the plot you've generated is the boxplot of those mean values.  This of course discards any information about the quality of sequencing along the length of reads, it's a very high level look at the data quality.

The reliability of single base calls is really not that great with nanopore sequencing, and that plot doesn't look particularly worse than other data that I've seen, but you can clearly see that reported confidence in the 2D reads is far better, as one would expect when you information from both strands.

Also, as a bit of advice, it's a good idea to use the name of the package your question relates to as one of the tags when you create a post.  That way the package author is alerted to your comment, rather than just stumbling across it.