Picard LiftoverVcf

I am having a problem with picard's LiftoverVcf.

I am trying to Liftover hapmap files (downloaded plink files from hapmap and converted to vcf using plink) from ncbi36 to hg38. I was able to do this with GATK LiftoverVariants. My problem came when I had to merge the hapmap.hg38 with some genotype files (that I liftover from hg19 to hg38 using GATK LiftoverVariants). I am merging them so that I can run population stratification using plink. I used vcf-merge but it complained that a SNP has different reference allele in both files: rs3094315, should be reference allele G (which was correct in the genotype.hg38 files but in the hapmap.hg38 files it was wrong). I also first tried to lift hapmap.ncbi36 to hg19 then to hg38 but the offending allele was still there. So I decided to try and lift the hapmap.ncbi36 using LiftoverVCF from picard.

1. I downloaded the newest picard build (20 hours old) picard-tools-1.138.
2. Used the command: java -jar -Xmx6000m ../../../tools/picard-tools-1.138/picard.jar LiftoverVcf I=all_samples_hapmap3_r3_b36_fwd.qc.poly.tar.vcf O=all_samples_hapmap3_r3_b36_fwd.qc.poly.tar.picard.hg38.vcf C=../../../tools/liftover/chain_files/hg18ToHg38.over.chain REJECT=all_samples_hapmap3_r3_b36_fwd.qc.poly.tar.picard.hg38.reject.vcf R=../../../data/assemblies/hg38/hg38.fa VERBOSITY=ERROR

Here is the run:
[Thu Aug 13 00:43:40 CEST 2015] picard.vcf.LiftoverVcf INPUT=all_samples_hapmap3_r3_b36_fwd.qc.poly.tar.vcf OUTPUT=all_samples_hapmap3_r3_b36_fwd.qc.poly.tar.picard.hg38.vcf CHAIN=......\tools\liftover\chain_files\hg18ToHg38.over.chain REJECT=all_samples_hapmap3_r3_b36_fwd.qc.poly.tar.picard.hg38.reject.vcf REFERENCE_SEQUENCE=......\data\assemblies\hg19\assemble\hg38.fa VERBOSITY=ERROR QUIET=false VALIDATION_STRINGENCY=STRICT COMPRESSION_LEVEL=5 MAX_RECORDS_IN_RAM=500000 CREATE_INDEX=false CREATE_MD5_FILE=false GA4GH_CLIENT_SECRETS=client_secrets.json

Here is the error:
Exception in thread "main" java.lang.IllegalStateException: Allele in genotype A* not in the variant context [T*, C]
at htsjdk.variant.variantcontext.VariantContext.validateGenotypes(VariantContext.java:1357)
at htsjdk.variant.variantcontext.VariantContext.validate(VariantContext.java:1295)
at htsjdk.variant.variantcontext.VariantContext.(VariantContext.java:410)
at htsjdk.variant.variantcontext.VariantContextBuilder.make(VariantContextBuilder.java:496)
at htsjdk.variant.variantcontext.VariantContextBuilder.make(VariantContextBuilder.java:490)
at picard.vcf.LiftoverVcf.doWork(LiftoverVcf.java:200)
at picard.cmdline.CommandLineProgram.instanceMain(CommandLineProgram.java:206)
at picard.cmdline.PicardCommandLine.instanceMain(PicardCommandLine.java:95)
at picard.cmdline.PicardCommandLine.main(PicardCommandLine.java:105)

1. I have no idea which SNP is the problem.
2. I do not know what T* means (does not seem to exist in the file).
3. I am new to picard so I thought VERBOSE=ERROR will give me something more but nothing more appeared.
4. Given that lifting hapmap.ncbi36 to hg19 then to hg38 produced the same erroneous reference allele I suppose lifting will not fix this and I will have to work with dnsnp to correct my file. Do you know how I can change reference allele in a vcf? Is there a tool for this? Is there a liftover tool for dbsnp?
5. As a side note I want to make picard work because I read that you will be deprecating the GATK liftover and will support the picard liftover (at some point in the future) so help with this tool will be appreciated.