The conical reducer sounds like an obvious option. But whatever you do, you're either going to have to do some re-engineering to make everything fit (and possibly reduce allowable pressure as well) or else bite the bullet and buy/fabricate the right item. If you put that conical reducer in, everyone that sees the thing for the next 50 years will wonder why it was built that way.
In the event that the thickness was considerably larger than the the offset, you might have other options. If the design pressure were actually very low, you'd have some other options.
Unless your company is fabricating this drum for it's own use, I would be hard pressed to believe that a purchaser would accept an extra cone section in their vessel.
Without the material/thickness/etc. information requested by weldtek, I would have to guess that in general, a new formed head may be less than or equal to the cost of the intensive cutting and welding required to resize the existing head.
Let me guess... you cut and rolled the shell BEFORE you strapped the head OD. Depending on design requirements and thickness, could you use a lap welded joint between shell and head?
You didn't say which diameter was correct, however, if the larger diameter is the one you want, or if not, but the owner could live with it, I would suggest cutting several feet out of the shell can(s) and splicing in a piece large enough to yield the correct diameter. Cutting back and adding a larger piece will enable you to avoid having two long seams just inches apart.
If you need the smaller diameter, the first thing I would do is contact the head fabricator to discuss the possibility of having him reshape the head(s). If you feel that none of these ideas will work, let us know, because there's obviously lot of talented people participating in this forum.
Just for curiosity's sake I'd like to know more of the specs of the vessel...length, orientation, supports, service...and particularly why the heads (~1") are twice as thick as the shell (~1/2")?
I can't see any customer accepting anything other than a "normal" shell correctly sized and attached to a correctly sized "normal" head, i.e. you fix the mistake so the vessel won't look like a "Frankenstein". While some of the suggestions are certainly "viable" options and structurally/code acceptable, I just can't seem to see some of them as anything other than "booty fab"...and I feel a customer would see them as the same, i.e. a cheap attempt to remedy a major mistake.
Check with your dished end manufacturer and get this re-pressed again to 3400mm diameter. Other option of making to shell ID to 3476 will call for approval from client/process, etc. I suppose 3400mm is correct diameter.
What are the design conditions which are causing disned end of 25mm thickness and shell of 14mm?