Drosophila- In situ hybridization problems

Our lab has been doing whole mount in-situ hybridizations on Drosophila embryos using random-primed or PCR-produced DNA DIG probes for years. In the last year, we’ve had serious problems with high background, and some problems with weak signal. After examing all components, we see that the most significant factor is the enzyme used for PCR! Most suprisingly, using Polymerase from Roche gives us background problems that are often the worst, and better results (but not consistently good) with enzymes like Super-Nova polymerase from NutriCyte (Buffalo, NY) and others. This is all in a system in which we’re: using Roche anti-DIG; pre-absorbing the Antibody with embryos; and blocking by methods that previously worked well. The background problems are not sequence specific, and arise with all probes.
Does anyone have suggestions about this problem? Or any general suggestions for improved ways of reducing background?