My advice is to try and avoid confrontations and disagreements in your formal meetings, just stick to the facts. I think you will find that over a long work career that there will always be people that you have to work with, or for, that are either incompetent and or just difficult to work with. That won't change, just their names change as you and they move along with your lives.
As you are working for a company making a product, you have the advantage of a formal relationship with the parts manufacture and certainly your local parts distributor that you purchase these fans from. A simple phone call or email asking if the 25khz PWM specification example is a fixed requirement or if 21.6 KHz or 28.8 Khz would work as well. I'm sure it would work fine, but in your case I would get it in writing via a FAX or an email if they say it would work fine at your other frequency choices.
In my work, before I retired recently, there were frequently questions coming up on specific instrumentation products we were using or wanting to use that couldn't be completely answered just by the product spec sheets.
Our local manufacturer's representative (salesman) would always search out and find the answers from the factory when we requested. That was one of his job duties to justify his commission and they would always try their best for us. Sometimes it could take a few days to contact the right person in the factory, but I was always able to get some kind of answer one way or the other.
So armed with data and facts you are in a better place to bring your case before your difficult person, and I agree it's better to go to him alone, at least give him a chance to change his mind before a formal meeting. Also it would be useful if you could estimate the costs to change the software to fit the 25khz spec (but it sounds like you already did that) so as to see the cost trade-offs involved. Professional engineering is often all about making trade-off decisions.
Other solutions could be hardware based. Someone already mentioned changing the micro's system crystal clock so as to be able to hit 25khz on the nose, as long as that doesn't require having to change other software timing issues if changed. The other method is to test the fan at your 21.6 KHz or 28.8 Khz frequencies and see if any of the other specifications change (speed, air flow, device heat, current draw, etc). Fit for service tests can always be useful if there is time. I've had many products that failed to meet their specifications and always trust real world testing over published specs, especially over things like accuracy specifications.
So there are always ways to proceed with technical disagreements ,but the best ways utilize as many facts as possible. Save the flame wars for these kinds of forums and you will most likely have a more successful career. I always tried to treat others as I would like them to treat me, even if some of them were/are ********.