How bad is bad - second plate of 2017 libraries

So to get right to the point, bad isn't all that bad, though it's not great. Here is the distribution of median read depth per clone for 2016 and the first plate of 2017 (just to refresh our memories).

Now here is the same distribution of median read depth per clone but adding in the second plate of 2017 libraries.

I added this version with each set of libraries marked with a different color. Red is the first set of libraries (SpFall2016), green is the second set of libraries (first plate of Sp2017), and blue is the third set of libraries (second plate of Sp2017). The reason there is that weird bar that is twice as tall as it should be about 2/3 of the way to the right, is that clone (DBunk_90), was sequenced in both plate 1 and 2 of Sp2017, so the way I am graphing it, it put that bar in twice and stacked it. Sorry about that.

We now have a lot more 2017 libraries falling out at a median read depth of 10 or below (N=45, N=5 before). But, only 4 of those has a median read depth of 5 or lower, and only 1 has a median read depth of less than 5. That particular sample was an individual we didn't phenotype.

As a side note, the D Mud clone I did manage to sequence does not look like a D. pulex, though maybe not completely like a D Barb/Oak either?

Here is the distribution of read depth per site for the D Mud clone. Notice the bimodal distribution with a peak at 0, which is what we would expect for mapping bias. 

This looks different than a DBunk clone for example, that has the same median read depth. Notice that you don't see a peak at 0 here.

But then the D Mud clone distribution also looks different than the D Barb clone distribution. Which is much more strongly bimodal. Though maybe that is just because the D Barb clone has a higher median read depth? Are the two distributions not that different, the D Mud one is just compressed?

New D Mud plot (same x-axis)

New DBunk 50 plot