I suspect the polygons are created as such large sizes, not because they are gaining the additional Z and M flags, but because they have millions of vertices! (Maybe some exaggeration there but.. perhaps you take my meaning.)
My recommendation would be to try downsampling the raster before you vectorize it. (By the way, never perform destructive edits on your original data!! Make a copy first!!)
Downsampling means if your raster grid corresponds to---for example---6 inches, you would translate it a larger cell-size, perhaps 1 sq. meter. Or maybe 1 acre. etc. etc. Fewer cells will mean fewer vertices, which will result in leaner polygon geometries. Just realize downsampling is costing you resolution. So you should be mindful to downsample into a reasonable resolution for whatever you're researching. If you're examining elk habitat, you don't need data in sq. inches. :) ...but if you're creating a cost-surface for pedestrian spaces, sq. feet or sq. yards might be reasonable.
If you're not interested in downsampling the raster first, you could always do the vectorization, then followup with a Simplify operation on your big polygons. This will take the polygons produced from the raster and reduce their densities, in effect accomplishing something similar---though I promise you the mapped areas will be different. The vectors from the downsampled raster will retain a "blocky" edge, while simplified polygons will probably look "jagged".
If I were you, I would downsample first, as I believe this would not only make the best use of your processor time, but also give you more desirable outputs.