A few points:
First of all, lots of people buy seeds, run though a pack or two to find a good "keeper" plant, then clone it to perpetuate it, or to fill up their growrooms. If you do this, and divide the cost of the pack over the dozens or even hundreds of clone plants you eventually generate from it, the per-plant cost is actually pretty low.
In terms of cost, remember you're taking months to do the grow, its taking up hours of your time, you're spending $X on bulbs nutrients and electricity, and you're going to end up with many ounces of herb at the end. Looking at it that way, even if you're paying $60-100 per pack of seeds per crop its not that big of an expense, relatively speaking.
*IF* you happen to be buying seeds of a true-breeding strain (aka an "inbred line") then you could just cross any male and any female to generate your own seeds. This is a viable option for many of the stable "old school" strains that have been around for a while.
Unfortunately, many if not most of the popular strains "of the month" are not stabilized hybrids, so in crossing two plants from a pack of seeds, you'd end up with any number of offspring plants with DIFFERENT characteristics. That's not desirable from a growing standpoint.
And this is really what's at issue here. As mentioned, the "problem" with creating your own strains is that doing so is highly labor intensive. If you want to create something NEW, you literally need to select from dozens if not hundreds of mature plants to identify the characteristics you want. Already that's beyond the capabilities of most small "personal" growers, who don't have the space, time, or inclination to do that sort of selection.
Next, to truly create a STABLE line, where all the seeds have similar and predictable characteristics requires at least 4-5 generations of repeated breeding to stabilize the genes, with 7-8 generations really being better. So a true quality breeding project to create and stabilize a new strain can easily take two full years, or possibly longer. Again, not something the average schmoe is going to want to take on.
For most growers, its actually much more time and cost effective to just pay $100 for someone else's genetics rather than try and create their own from scratch.