The bibliography is interesting and useful but not immediately evaluable, and seems mainly Canada-focused. The Stop It Now/D2L data is very clear up front that it relates to prevalence and not incidence, a distinction that matters for reasons to which we'll return in a moment. The UNH data substantiates some level of decrease since the early 90s, as does every other source, but disagrees with itself on magnitude in different places and comes out of one researcher's shop, which does not invalidate it but does mean I'd like to see someone else produce something similar before I weight their novel results too heavily.
I guess maybe we have differing definitions of "reliable," but I think you and I have clashed on this subject before. I don't blame anyone for assuming that a victim of such abuse would tend to overestimate its prevalence, and I try to watch out for the tendency in myself, but what's all the anger about?
The difference between prevalence and incidence matters because the former tells us how many people have been abused as children, and the latter tells us how frequently children are abused over time, which is the thing we're actually talking about. How we count multiple instances of abuse against the same person confounds everything, and intuitively I would expect prevalence by victim (ie for a given victim how frequently they themselves were abused) to follow something like a power-law distribution, but in general if we just take the headline number of "X total kids" it's hard to see how that gets us to a nuanced understanding of incidence rates that lets us talk knowledgeably about how and to what extent the shape of this problem is changing over time.