In other words it is more a matter of personal tastes rather than a real issue.
Well, everything is a matter of personal taste. I gave my personal opinion, but I doubt you would find it to be the prevailing opinion among BlitzMax users. A lot of people are serious about developing for Linux and/or Mac, and the fact that the CLI is far from complete on those platforms is not going to be a "small irritating issue", nor is the fact that you have to go use SharpDevelop instead of Visual C# if you even want to try and write cross-platform programs. My issues with it are personal taste, sure, but then I did also say that I haven't done enough programming with C# to be able to say categorically that there aren't other huge problems with C# or indeed other huge benefits with C#. It could easily be much better or much worse than I have found it to be when you start getting deep into a big project.
You would agree that it was much better to expose Blitz functions to C# or C++ ,same as they did some years later, rather than wasting a lot of time for developing a new programming language
Well, no, obviously I wouldn't agree. I'm not using the B3DSDK now, nor would I have done if they had released it three years earlier. Back then, I had no idea that a more powerful language would benefit me so much, so there's no way I'd have started using B3D's graphics engine in C++ or C#. And now I have TV3D which is far more powerful. Evidently, I am using BlitzMax, so for me it was obviously much better that they did what they did.
Or were you asking me if I agree it would have been better from BRL's point of view? I really couldn't speak for them, but I do note that the B3DSDK forum is very, very quiet, so I wouldn't have thought sales have been soaring. Then again, as you say, they could have done it three years ago if they hadn't developed BlitzMax, so perhaps it would have been much more popular then. There were less alternatives and it was less dated. So possibly.