No idea if this is any better but from PacBio itself: https://github.com/PacificBiosciences/pb-assembly

We've had good success with both Flye and Canu assemblers; however, if your reference animal is highly heterozygous or if your input sequence is highly contaminated, you'll get fairly poor results in your output assembly. Also, lack of sufficient coverage (targets should be > 50X coverage for a diploid, homozygous mammalian genome) can be an issue.

My recommendation is to try using Merqury or KAT to assess the quality of your input sequence reads compared to your assembly. If you notice a huge error band (the leftmost bar) or the lack of het/hom kmer peaks in your kmer spectra plots, then you have some issues with contamination. I'd recommend running reads through Centrifuge or another tool to remove the contamination first before assembling.