I like this question. Hope the tutors can offer guidance.
I've personally got a lot of thoughts in my mind now and seem to be thinking about extreme scenarios:
1) Yes, the cost of providing the guarantees will increase because of the reductions in asset shares, BUT;
2) When a company is offering a guarantee, it will take a view as to the average expected loss it will make as a result of this guarantee (so for eg out of a 1000 stochastic projections, 300 scenarios might show the guarantee bites, and the average of these 300 will be the cost of the guarantee). This loss to the company is in other words a profit to the policyholder.
3) Therefore, based on (2), the intention of reducing the asset share is to compensate for the loss to the company or to charge the policyholder for the profit they are expected to make. Since this is iterative, in calculating for this charge you will want to stop at some point naturally where the balance between the cost of the guarantee and the reduction in asset shares is at an optimal position otherwise you'll end up reducing asset shares to zero and using up the estate to fund the whole benefit under the product.
4) The only missing portion to balance the equation then is how the charges we apply are actually used; ie maybe these are used to purchase put options that give a benefit close to the gap between the asset share and the guaranteed benefit; or this is converted into a higher premium that the policyholder can pay which after accumulation, reaches the guaranteed benefit etc; or again just simply used as a reserve for the anticipated losses to come for the company.