I see what the problem is, its your phobia to comparisons. The important point is that advantageous traits are wiped out less frequently than non-advantageous traits, not neccessarily in every specific situation but less in general. If we accept that most mutations are wiped out regardless of their beneficial status then it still makes no difference to the fact that some survive and of those beneficial traits tend to come to predominate. They are generally preserved
more than non-advantageous mutations.
I think deleterious mutations in germcells killing them before they get to form a zygote should be included yes, because advantageous mutations which apply to germcells getting to the zygote are also included.
Why should this be the case, the thing which makes a certain sperm for instance be successful in reaching the egg need not be something which is carried in the genetic material within the sperm and certainly not as a novel mutation. But this is beside the point. I think you are getting confused between two concepts, one of counting all mutations lost due to never forming a zygote and another of beneficial mutations being lost due to the presence of embryonic lethal defects in the embryo.
If you are counting every mutation present in every germ cell regardless of its contributing to an actual organism or not then I doubt anyone could argue that, given that assumption, most mutations are lost before they get a chance to propagate. These are never mutations which constitute part of the population however and certainly not part of the breeding population.
What it comes down to is a strawman argument, by putting forward a misrepresentative meaning for the phrase 'advantageous mutations tend to be preserved', i.e. that any such mutation ocurring in a DNA sequence in any cell will tend to be preserved or at least in any cell which might contribute to the next generation, you hope to make it appear as if there is a problem with the way it is presented in the context of natural selection, which is not the context you are putting forward.
By far the vast majority of the mutations wiped out in such a way never get a chance to be expressed as a phenotype and can hardly therefore be expected to be a suitable substrate for natural selection.
TTFN,
WK