Due to pages ( characters ) constraints of a journal, I am forced to 70% of results and deeper description of methods, quality assurance and characterisation techniques submit as the additional information paper to the original. It will not be printed, but it is accessible for free on the website. My advisor is telling me not to put so many, it is bad and I am risking to be denied by editors. Why? Also I asked on some other forums, other people also tell me that if you refer a lot to SI your paper will likely be denied. Is this right? Advice that I got is to wait until editors or reviewers demand to see them additionally after submission.
There's a difference between putting a lot in the SI and referring a lot to the SI.
If you have a lot of results (for example you created a lot of new compounds) your SI has to be big to fit all the characterization data. I'd say it's also good to put more detailed version of the experimental part in the SI. The version in the paper should make clear what and how you did something, the SI could contain a version that's actually usable as a protocol. Length shouldn't be an issue here.
Referring a lot to the SI is not that great. If you have to refer to the SI a lot, this shows that the paper is not streamlined, you're not focused on the main findings, and you're making people read two manuscripts at once if they want to go into any detail. The paper should stand on its own without the SI. Tthe SI should be something extra, not required reading.
In your case, it could be that you're needlessly referring to the SI in the text regarding characterization and quality control. Those references do not need to be there, you can put the summary of the SI at the end, stating that the SI contains all characterization and quality control data (and protocols).
As a good example you can look at one of the latest papers from the group of Phil Baran, active in natural product synthesis and new synthesis methods. http://pubs.acs.org/doi/abs/10.1021/jacs.6b13229 has a SI of more than 900 pages (not online until it's out of the "just released category), and there's not a single in-text reference to the SI. A slightly older example has a 127-page SI that is online at the moment: http://pubs.acs.org/doi/abs/10.1021/jacs.6b00250