confusion about genotyping and imputation

Question

[Zhenyu Zhang]

I have a bigger question here. Please let me know which part of my understanding is wrong.

1. My understanding of array-based genotype calling algorithms are mainly based on cluster, which can only call genotype as AA/AB/BB. Then you have to have the probeset/SNP table to tell which two bases are here, and even if you do and know which two bases are here, you would have no way to distinguish which base is A and which is B. The only thing you can do is to guess based on minor allele frequency; however, this could be very wrong when maf gets close to 0.4~0.6.
2. If my understanding of 1 is correct, and this flips are not big problems for GWAS directly; however, imputation algorithm need to know exact haplotypes you have in the region; with these flips, you expand you errors.
3. Is the answer that because even there are flips, there are likely to be clustered together in the genomes. And also maf 0.4~0.6 provides less information for imputation compared with maf 0.1~0.2, so the errors are there, but are not very extensive. Or some parts of my understanding about genotyping and imputation are completely off?