When you reference a paper within whatever your context may be, you are possibly doing several things. First, you may make claims and use other authors names and reputation in support. Second, you may take "facts" from a paper and propagate these facts through yours.
Nothing wrong with that? Not generally. But, what happens if a paper makes a claim that is not at all well supported by the study? You run the risk of propagating errors so that when somebody uses your paper as a reference the original paper is still further away and after a few such iterations the source may be completely forgotten. There are many instances where either errors have been propagated or where "truths" have slowly been misquoted so that they turn into errors. This is clearly not what we want in our papers.
I would therefore say that one needs to (critically) read a paper enough to make oneself sure that the facts can be trusted and that no misinterpretation has occurred in the paper to be referenced or earlier. Hence relying on, for example, other authors references is a very weak link in the chain. One has to try to back-trace vital information as much as possible. Misunderstandings may not necessarily be born out of malice but just by oversight, but the end result is still the same. To therefore, for example, simply gloss over the abstract and use whatever seems to support some idea or vice versa is far from satisfactory.