"a value-free sociology is impossible to achieve and would be undesirable anyway." How valid is this assertion?

I'd say it's quite valid, and applies generally to all social sciences.

While it is important to make careful, accurate observations and avoid bias as much as possible, the idea that we could ever eliminate all value judgments or even all biases is unrealistic. We do the best we can. Simply by deciding to do research in the first place, we are effectively acting on the assumption that the knowledge we will acquire is valuable.

Moreover, it can be dangerous to think that we are eliminating all biases and doing "value-free" research; generally what happens then is that we make implicit, unexamined value assumptions (one that comes up a lot in economics is "maximizing GDP is good"). It's better to instead acknowledge our value assumptions so that we can examine and question them.

On the other hand, this doesn't mean we can just freely inject our own values wherever we want---that would clearly lead to biased research. Instead, we clearly state our value assumptions, and focus on the primary goal of getting objective, accurate observations. If we study a behavior we find morally objectionable (infanticide occurs in many cultures, for example), we try to set that aside when studying it, in order to better understand the motivations behind it. At the end we may still decide that it is morally wrong and want to get rid of it---but by understanding it better we will be better equipped to recommend policies that will actually be effective in doing so.