the difference between linear & point filter

Yes. If you sample a value between two pixels, you will get the average of the two values, whatever the render target ratio dimensions are.

 

For blur filters I would use a linear filter since that'll add more blur for free.

 

Perhaps "weighted sum of texels" would be a better way of expressing it?

 

Yea, and if you align the pixel centers and texel centers properly, you will only get weights of 1 and 0, hence one pixel is used completely, and its neighbours aren't used at all.
So you effectively get a pointsample.

What I got from the opening post is that he was asking whether the linear filter was turned off altogether when the texture and rendertarget are the same size (1:1 ratio). This is not the case. The linear filtering is still performed, it just yields the same result as pointsampling when your pixels and texels are aligned exactly.

 

Indeed...

 

I would have thought "Weighted average" was the best way to express it.
(Not native english though.)