Other native proteins required for cell growth in your competent cell have Histidines which could bind to a nickel column and you can't prevent that. It is also possible that those proteins are binding to your protein and being eluted.

When you express your protein in competent cells, you have no idea where the polyhistidine tag actually is located in terms of solvent accessibility. If the C-terminal tag is tucked into the interior of the protein, it not solvent accessible so it won't bind to the Ni-NTA Agarose beads very well. If it is solvent accessible, it will bind well. If you put the polyhistidine tag on the N-terminus and it bound to Ni column more tightly then solvent accessibility of the 6X-His tag would be why. That is not something you can predict or know in advance.

Some scientists actually unfold their protein with urea so that their His-tagged protein will have an accessible display of Histidines to binding to Ni-NTA tightly. That is dangerous, though, because although your protein bound tightly to the beads, no you need to refold your protein either on or off the column before Imidizole elution and now you have to verify that it is refolded and behaving like a properly folded protein would.

@Sebastian Schmitt provides lots of good advice, you haven't mentioned what your sample loading is vs your column volume?

IMAC is a poor affinity purification system, if you have too much capacity you invite non-specific binding and end up with a dirty prep. I normally advise under capacity, as this allows your tagged protein to compete off non-specifics. However, as you are working with membrane proteins this is not commonly an option.

Have you pooled your wash step? If you dilute and rebind, you can sometimes end up with a better prep than the "purified" material that bound to the column with the non-specific proteins.

Thank you all for your suggestions. To answer Sebastian Schmitt