While I was studying an infant's transcript, I realized that he deleted the [l] sound in "alma" [alma], a word in Turkish meaning "do not take". When he deleted the sound, the word became [a:ma]. There is another word; "ama" [ama] in Turkish meaning "but". These two words become similar unless he changes any sound. However, the infant utters them differently after deleting the sound. I wonder whether the infant delibaretely changes the first sound of "ama" meaning "do not take" to show that they are different words. Does he mentally try to show that he knows that they are different words and he cannot only utter the [l] sound?
The first thing to do here is separate out a child's knowledge from their production. The most well known and regarded example of this comes from J. Berko and R. Brown (1960). "Psycholinguistic Research Methods". In P. Mussen. Handbook of Research methods in Child Development. New York: John Wiley. pp. 517–557. They give this example of a child interacting with an adult. The child has a toy fish and says /fis/, the adult then replicates the child's pronunciation and says 'is that your /fis/?' The child says 'no, it's my /fis/" (and is rather ticked off at the idiocy of the adult). When the adult finally asks if it is a /fish/ the child accepts the question.
What this example shows is that even though children often produce incorrect forms at various stages of their development this doesn't mean that the child can't perceive the difference between what they are saying at the correct form. So even though it may sound like the child in the example you give is merging two forms it is very likely that they're aware of the difference and it will resolve when they've got their tongue around that tricky consonant cluster.