I understand your frustration at this seeming inconsistency. I think part of it is that fundamentally, arguments about ID and whether they work have a philosophical bent to it that many scientists are not trained to recognize. Thus we have scientists attempting to refute ID by arguing that nature has “bad design”, which in my opinion is equally as misguided and confusing as ID arguments. So your misgivings are somewhat justified.
That being said, the nature of ID arguments are such that they justifiably lend themselves to this two-minded rhetoric. Firstly, this is because ID advocates often do bring forth scientific arguments asserting that evolution cannot explain certain feature X. Such claims can be engaged with in purely scientific terms, by showing that evolution can indeed plausibly explain X.
Secondly, scientists in principle are not opposed to the term “design” (or any new term, for that manner) as long as it is defined in a rigorous and measurable manner, which would make it a legitimate scientific question. However, in practice it turns out that the definitions of ID proponents are not rigorous enough for most scientists, which is why, for example, even after a very technical discussion, Josh concludes in his dialogue with ID theorist Eric that Algorithmic Specified Complexity is ultimately still a negative argument - “if we cannot adequately model a pattern it must be the work of an intelligence.”
It is at this point that the scientist critic of ID points out that ID is not legitimate science because it can’t define its concepts rigorously, despite the layers of sophisticated information theory that frame it. Thus this two-tracked criticism of ID is justified because a huge part of the dialogue is probing precise definitions, which is done at least in part by first attempting to engage with ID concepts assuming they are legitimate science.