I can't help but reiterate here Serre's great maxim: "Precise yet informal". Striving for this goal in mathematical writing will avoid your looking unprofessional, but will make you look more human at the same time. It seems that mathematics that does this best is very `professional' professional (I can't help but think of Milnor's style, here).

Regarding writing that appears purely to guide the reader between theorems: as far as the guidance ALSO is precise yet informal, there should be no foul. It is when these comments are not really true...but close to true...that they should be omitted or left to the talk. Too many "sort-ofs" should be avoided in the body of a mathematical paper. Serre's maxim seems a good guidepost for this, too.

Comments to supplement original answer:

Also, as long as mathematical papers are valued for being interesting it seems that there will always be a bit of 'ought' along with the 'is'. Otherwise, one could just rattle along formally proving theorems without regard to how they will be received. This is evidence of the human part of mathematics that stands in opposition to the dispassionate precision of formal proof. Elegance is far from formal perfection, and is perhaps closer to human utility or aesthetic sense. Somehow, Serre's maxim expresses something important about mathematical life and work.

I can't help but reiterate here Serre's great maxim: "Precise yet informal". Striving for this goal in mathematical writing will avoid your looking unprofessional, but will make you look more human at the same time. It seems that mathematics that does this best is very `professional' (I can't help but think of Milnor's style, here).

Regarding writing that appears purely to guide the reader between theorems: as far as the guidance ALSO is precise yet informal, there should be no foul. It is when these comments are not really true...but close to true...that they should be omitted or left to the talk. Too many "sort-ofs" should be avoided in the body of a mathematical paper. Serre's maxim seems a good guidepost for this, too.