Everyone wins with this approach. Sure, those lazy bums didn't consider cases X, Y, and Z, but then no one else had solved the problem regardless of whether or not they handled X, Y, or Z. That's how A, B, and C's results got published!
You're building on their work. You look better for acknowledging it rather than trying to knock down those who came before. They are likely to be respected quite a bit better than you are. Also, researchers A, B, and C are the most likely candidates to be reviewing your paper. So there's that.
This is, of course, orthogonal to the issue of unreproducible experimental results.
Sometimes, however, there is no easy way to avoid saying what is missing. As an example, consider the case when the published research, without giving any explicit justification, does not address one aspect of a problem that the other researcher working in the same area finds important. But even in that case, you should not fail to acknowledge the work that you are building on.
As another comment suggests my interpretation of 'negative citation' is probably less strict. I considered negative even the case of pointing out a shortcoming in a polite way, without questioning the validity of the work in a confrontative manner.
Even when a previous paper was completely wrong (say a chemical structure was incorrect) very rarely do follow-up papers make a negative comment. They usually say "our results differ and this is why we think we're correct".
I think a part of it too is that it's not hard to be fooled by your own results. Even a relatively innocuous error or failure to notice something in the date is really easy to do. I think follow-up papers are less critical since scientists can understand why the original paper was wrong.
Improving on engineering is nonconflicting. The existence of an excavator doesn't refute the shovel, and so on.
A disagreement in, say, CS would be, "Foo Barley [42] mistakenly estimates an upper bound of O(n log n) on his algorithm; here we prove, in fact, a quadratic lower bound! Moreover, the algorithm calculates produces results in several cases that Foo Barley doesn't consider, presenting it as general."
How much of that do we see?
I used to be in the NLP group of a good school on the east coast. I was trying to reproduce an algorithm from a paper that won Best Paper from a good school on the west coast. A postdoc from that school said, "oh stay away from that algorithm. they basically cooked the results of that whole paper by cherry picking data. it doesn't really work at all".
At that point I dropped it and never thought about it again -- there is no career value for a phd student (the engine room of research) in "outing" bad research. I was just trying to get some baselines to compare my own work against. On the flip side, getting best paper award at a top conference is the kind of thing that sets you up for faculty interviews at you choice of university.
All that said, academia and the scientific process tend to work well, in aggregate and over time. I think a lot of times people with juicy or disgruntled stories speak more and get remembered more, and it casts academia in far more negative a light than reality. Most of the time, science works well.. just in a way that doesn't produce stories worth posting to boards like HN. So that's my counterbalance.
EDIT: That may have to do something with peer reviews: they kind of prevent the publication of work that will be openly disagreed with later on.