The problem lies in the regular expression you are using with "$GeneDetailrefGene".

See below:
What you are essentially saying is that in the string $GeneDetailrefGene, look for one of the following:

a) a dot character (".") followed by 1 or more digits [0-9] followed by any one of the characters "+", "*" or "-"

Code:

\.\d+[+*-]

== OR ==

b) a non-digit character

Code:

\D

whichever comes first.

If either a) or b) is followed by 1 or more digits [0-9], then capture those digits and set them to $transcript.

The "whichever comes first" part is crucial here.
If you have a regex like this: "A|B", Perl will match either A or B, whichever comes first - reading the string from left to right.

So, when your program sees this value for the $GeneDetailrefGene:

Code:

NM_001134408:exon3:c.415-43A>G;NM_001134407:exon3:c.415-43A>G;NM_000833:exon4:c.415-43A>G

it matches the "_001134408" to "\D(\d+)" and therefore sets $transcript to "001134408".

For a clearer explanation, the red part of the regex below matches the red part of the string $GeneDetailrefGene. Ditto with the blue part, which is eventually assigned to $transcript.

Line 1:

Code:

NM_001134408:exon3:c.415-43A>G;NM_001134407:exon3:c.415-43A>G;NM_000833:exon4:c.415-43A>G
/(?:\.\d+[+*-]|\D)(\d+)/

Line 2:

Code:

NM_003239:exon6:c.927-1G>C
/(?:\.\d+[+*-]|\D)(\d+)/

Line 3:

Code:

NM_001879:exon12:c.1442-5C>G
/(?:\.\d+[+*-]|\D)(\d+)/