It's bad but...
Realistically your investment in website availability should largely compare to how much of your business depends on your website. If you're new to the web with a business that does not rely on the web, then not so much.
If you're running an e-commerce site it's much more important that people be able to reliably access your site, that your site works in terms of valid security certificates, and that they receive their orders in a timely manner.
If your site performs a service such as look-ups, searches, social media, or maintains a repository of information people may rely on, then it's probably not as important because those people will definitely continue to come back even if they experience problems. Especially if you're providing a subscription based service. That being said if Facebook was down for 18 days straight people would do something else.
If your site is for information purposes only (a web presence) then this isn't really that much of an issue at all (provided the days aren't consecutive) since most of that information has likely been mined and spread around the internet in various forms. It's likely been cached by major searches as well. People will call and they will still e-mail. They might also look for a replacement vendor though if they think you're another victim of the recession.
If the site is a microsite for a zero-day launch advertising promotion and the site is down on the day of the promotion, then this is a catastrophic failure.
In a similar situation one of my clients sent out an e-mail blast to their main 20,000 name list the day their domain registration wasn't renewed. The site was down for about 2 hours while they tried to figure out what was going on. They contacted me and I traced the issue quickly (remotely), but their IT personnel weren't being copied on the e-mails from their registrar (management issue). Their main e-mail services went down, bounces didn't come back to the proper server and many of the people on the list unsubscribed (or did so shortly thereafter). They ended up on an e-mail blocklist since they were sending the e-mails from the same IP pool as their main mail server. It looks like during a postmortem it was a $30,000 mistake based on lost potential, downtime, and the marketing investment required to rebuild the list. It would have been worth it for them to have had some sort of uptime check fail safe in place. When you host with a shared host or even a cloud-based service there is the potential that DNS could go down causing this same issue. If your website IP can't be resolved then you could lose that traffic, email, etc.
"What impact does a not so good availability (let's say 95% - 18 days in a year) have in User Experience?" Let's say the next time you go to an ATM to draw money out of your account that you consider the possibility of your bank not being available 18 days out of the year. What if they were consecutive? What if it only happened on days with high traffic (like payday).
Most hosting providers these days provide some sort of "shared cloud hosting". It's not a requirement to use something as expensive as collocation, but cloud hosting has its own drawbacks. Also many hosts can replace a server in a about an hour now and restore a back-up image from a SAN in the event of a power supply failure, a hard drive failure, or a bad battery back-up. If you're in a cloud being hosted by a company like Amazon, then you will have no say when your site goes down unless you're a major monetary influence. They will focus their resources on the much larger fires.
SEO - if it's your lifeline...
One of the other things to consider is if your visitors can't access your website then neither can the search engines. Search engines will crawl a site at random several times a month. If the site is down they will likely crawl when the site is up again. If there are issues with the site not being available for any reason, the search engine will likely send your traffic from searches to someone who can actually serve the pages.