You're referring to the "generate-recognize" theory of memory recall. It's actually an interesting theory, which states that you can't actually recall memories directly -- at least ones from more than a few weeks ago -- because your brain does not store that much detail about them.
Instead, the theory states, memory recall is a rapid-fire, iterative process of "generating" details of a memory, then checking whether it matches the original. If it does, then the detail is 'recalled' and the process moves onto other details, in this way hopefully producing a reasonable reproduction of the original memory. There is some experimental evidence of this theory, but as far as I understand it, the theory is not entirely accepted. In my opinion, it's probably part of the story but not all, perhaps as an 'error correction' mechanism that fills in gaps but not for producing the entire memory.
What you're referring to about false memories is somewhat different and much better accepted. There's a lot of evidence that shows that memory recall is a destructive process: after recall, a memory becomes labile and must be reconsolidated into long-term storage. Essentially, recalling a memory destroys it and it must be re-encoded as a new memory. In that time while it is labile, details can be added or removed and changes can creep into it from your own thoughts or the suggestions of others. The changed version is usually indistinguishable from the original.
