I do not beleive in the existence of junk DNA. Nature is a very conservative and economical lady who will not waste time, energy and resources to produce junk.

At least some non-coding DNA may be simply serving a space filling function. That is, it may be maintaining required minimum physical distances between coding regions, or certain space requirements around centromeres related to physical constraints on the process of meiosis and replication. That sort of functional role is very difficult to determine experimentally, but there is evidence for it.

However, I would add that is not entirely inconceivable to me that some non-coding DNA serves no significant purpose at all with regard to the host organism at least. Nature is efficient, so that if an insertion ends up serving no function basis, but it also does no harm and does not impede any other necessary function, then it may well persist.

So while I agree Nature is efficient, that does not necessarily mean it finds a use for everything. If we're going to anthropomorphize, I'd say that Nature as a whole, when it comes to an organism's genome, tends to follow the caveat of "first, do no harm". So a sequence that inserts itself into a hosts genome but which has no significant effect on fitness may well be "allowed" to persist, relative to the cost of actively excising it. Or a gene with multiple copies becomes non-functional due to a mis-sense/non-sense mutation, but whose loss is not really significant due to the redundancy of that gene, that now non-functional copy may persist as it bears no signficant cost to the host. That sort of thing may explain why so many pseudogenes or remnants of pseudogenes can be found in many genomes, and why at least some of these seem to persist for very long periods of evolutionary time.

I think this issue of non-coding genomic material is again a situation where vast generalizations server no real purpose (i.e. it's all junk, or it's all not junk). Much of this material may very well be meeting all sorts of necessary functional requirements that we just do not understand yet. But some of it may also really be doing nothing of any significance or consequence, and perhaps just represents part of the "cost" of evolution and fitness in a constantly changing environment.