Genetics lab question: unexpected phenotypes in male parental backcross

Hi,
I crossed true breeding AB+ females with a balancer chromosome (don't recombine) to true breeding ++c males (A,B,c on same autosomal chromosome) and got F1 apparently heterozygous A,B,+ phenotype. I did a male parental backcross of these F1 males to true breeding ++c females, and got 4 offspring with the unexpected phenotype of A,+,+. This shouldn't happen since male Drosophila don't recombine, and it throws off my data analysis for determining . What could explain this (other than contamination)?

i.e.
AB+//AB+ f  x ++c//++c  m   apparently A,B,+

if [AB+//++c] m  x ++c//++c  f A,B,+ and +,+,c expected and observed, A,+,+ not expected and observed (rarely)


Thanks!