There is some literature which shows that all start codons code for methionine. However, in the standard genetic code, the alternative start codons clearly code for leucine. Does that mean these codons will code for leucine when they are encountered during translation (after start codon has been initialised and translated).
I commented that this was a duplicate, but reading the question more carefully you seem to be asking something slightly different.
In the context of a 'start' these codons will be recognised by fMet-tRNA and a formyl-methionine will be inserted as the first amino acid. Subsequent occurrences of the same codon within the open reading frame will be translated normally (e.g. GUG > valine).
The use of GTG as an initiation codon in the E. coli lacI gene
In this paper
Frottini et al. (2006) The Proteomics of N-terminal Methionine Cleavage. Molec. & Cell. Proteomics 5: 2336-2349
the authors report assays of E. coli methionine aminopeptidase with model peptides showing that when Lys is the 2nd residue, Met removal is very inefficient. They also show that when Pro is the 3rd residue, Met removal is very inefficient.
This explains the fact that when the lacI repressor protein was sequenced:
Beyreuther et al. (1973) The amino-acid sequence of lac repressor. PNAS 70: 3576-3580
it was found to have a Met residue at its N terminus (sequence Met-Lys-Pro-).
However, when the lacI gene was sequenced:
Farabaugh (1978) Sequence of the lacI gene. Nature 274: 765-769
the corresponding DNA sequence was GTG AAA CCA, demonstrating that the N-terminal residue is encoded by a GTG(Val) codon. LacI residue Val23 is encoded by a GTG codon, demonstrating the normal use of that codon in the body of the mRNA.