I just heard a comment by Braddock of Lovesystems that was brilliant: All that your brain does when you ask it a question is hit "search" and return the first hit it finds. So be careful how you phrase your question.
Say you just arrived at work, and realized you once again left your security pass at home. You ask yourself, "Why do I keep forgetting my security pass?"
If you believe you are a rational agent, you might think that you pass that question to your brain, and it parses it into its constituent parts and builds a query like
X such that cause(X, forget(me, securityPass))
and queries its knowledge base using logical inference for causal explanations specifically relevant to you and your security pass.
But you are not rational, and your brain is lazy; and as soon as you phrase your question and pass it on to your subconscious, your brain just Googles itself with a query like
why people forget things
looks at the first few hits it comes across, maybe finds their most-general unifier, checks that it's a syntactically valid answer to the question, and responds with,
"Because you are a moron."
Your inner Google has provided a plausible answer to the question, and it sits back, satisfied that it's done its job.
If you instead ask your brain something more specific, such as, "What can I do to help me remember my security pass tomorrow?", thus requiring its answer to refer to you and actions to remember things and tomorrow, your brain may come up with something useful, such as, "Set up a reminder now that will notify you tomorrow morning by cell phone to bring your security pass."
So, try to be at least as careful when asking questions of your brain, as when asking them of Google.
The nature of self-defeating behavior is to be self-sustaining. Or to put it another way, our problems usually live one meta-level above the place we insist they are. (Or perhaps one assumption-level below?)
IOW, the arguments we repeat about why we can't do something are correct, if viewed from within the assumptions we're making. The trick is that at least one of those assumptions must therefore be wrong, and you have to find out which ones. The original NLP metamodel is one such tool for identifying such assumptions, or at least pointing to where an assumption must exist in order for the argument to appear to make sense.
There are at least a couple ways you could end up cycling, that I can think of. One is that you're not actually connecting with your near-mode brain about the subject, and are thus ending up in abstractions. Another is that you're not placing enough well-formedness constraint on your questions. At each level, you have to imagine that you already have ALL the things you wanted before.... which would make it kind of difficult to cycle back to wanting a previous thing.
In other words, the most likely cause (assuming you're not just verbalizing in circles and not connecting with actual near-mode feelings and images and such), is that you're not fully imagining having the things that you want, and experiencing what it would be like to already have them.
This is a stumbling block for a lot of techniques, not just Core Transformation. The key to overcoming it is to notice whether you have something preventing you from imagining "what it would be like", like that you think it's unrealistic, bad, or whatever. Noticing and handling these objections are the real meat of almost ANY mindhacking process, because they're the "second meta-level" issues I alluded to above, that are otherwise so very hard to notice or identify.
If you don't address these objections, but instead just plow through the technique (whether it's CT or anything else), you'll get inconsistent results, problems that seem to go away and then come back, etc.
(NLP sometimes refers to these things as "ecology", but relatively little time is spent on the subject in entry-level training. It's something that you need lots of examples of in order to really "get", because the principles by themselves are like saying you can ride a bike by "pumping the pedals and maintaining your balance". Knowing it and doing it just aren't the same.)
Sadly, NLP practitioner certification at best means that you learned some REALLY basic stuff and were able to do it when supervised, and while doing it with people who are receiving the same training at the same time.
That is, NLP certification drills are done by trainee groups, who thus already know what's expected of them, which means nobody gets much experience of what it would be like to walk somebody through a technique who didn't receive the same training!
Not actually what I said: it's about not allowing ourselves to feel good unless certain conditions are met. Or more precisely, our brain's rules about feelings are not reflexive: if you have a rule that says "feel bad when things don't go well", this does NOT imply that you will feel good when things do go well!
And, you will actually be better off having rules that tell you to feel good even when things don't go well, because bad feelings are not very useful when it comes to motivating constructive action. They're much better at telling us to avoid things than getting us to accomplish things.
(By the way, another common cause of self-defeating behavior being self-sustaining is that we tend to filter incoming concepts to match our existing frameworks. So, where my phrasing was ambiguous ("allow ourselves to feel certain things"), your brain may have pattern-matched that to "feel what we're feeling", even though that's almost the opposite of what I intended to say. The "certain things" I was referring to were feelings like the Andreas's notion of "core states": things that most of us aren't already feeling.)
The practitioner I went to was specifically certified in Core Transformation, not just NLP.