I think that you can be pleased with the overall result given the equipment that you are using. You are fighting against effects which are unbeatable without cheating and compromise. If you want substantially better you need better equipment or you need to "cheat" or both.
There are several apparent affects at work here.
As Matt says, there are substantial JPG compression artefacts. The original file from imgur was 750 kB suggesting it had been compressed substantially somewhere along the line.But, this is probably not a major contributor.
Consider the marked up version below.
You can also look at the marked up imgur version here and
As marked up version here - click icon 2nd from right above photo to "download original file".
- The target is just too big for the available pixels!!! - it's a magnificent view but
The 550D in full song produces a maximum 5184 x 3456 = 18 mp image and the imgur copy was 3888 x 2592 = 10mp so it's about 45% of the resolution of the original area wise or about 75% linearly. Even given another 33% you just haven't got enough pixels for the details. Here's an example where the bloated D800 sensor almost makes sense.
Camera needs to be set to produce max-everything. RAW if using RAW. Best possible jpg if not. EXIF or at least core data for this particular image would be useful.
See note G on image - the streetlight has been blown up to show the detail. That's a 30 x 30 pixel original as it came to me. There is not enough detail there for the light to be expected to be much sharper than it is.
You may be getting visible diffraction effects. The tasteful starring on the street lights is often a sign of small aperture and possibly diffraction. You say f16 which seems probably about marginal for such effects.
The sensor is saturated in luminance (highlights "blown") and so also in some or all 3 colour channels in many areas.
You cannot expect "genuine focusing" of original material if the highlights are fully saturated and detail is destroyed.
This is an unavoidable result if you want to render a picture with such wide dynamic range and achieve an eye pleasing result without resorting to HDR, tone mapping or magic in general. Notes A. B. C, E show histograms of small areas.
The light at A has a nicely blown spot of 100% white ijn place of the light. The brain accepts that without much complaint.
The light at B also has it's max level fully saturated - red channel shown here - but has a very large component of output across the range. (What the histogram shows depends on how much surrounding dark gets included - but it's clear that "tone mapping" or HDR or in camera local contrast management (DLighting etc) would render this better if violently enough applied.
I included C to show that a line of lights with surrounding moderately lit areas has much full saturation even though this is not necessarily obvious at a glance.
Where's Wally / D ? ?
E is on the river reflections - the brightest spots are saturated.
I looked at F to see if some sort of filtering might help the lights. Blue is well down but red and green are well saturated in the centre.
Overall it's a nice image.
It would arguably tolerate a somewhat reduction in total exposure and this will help some of the more blown highlights - but only somewhat.
This is an example what HDR is made for (an opinion, feel free to disagree :-) )- tastefully used, possibly with manual control, it will probably add much - but the effort may be severe and many will not notice.
Selective contrast or brightness control would also probably help - depending on how the camera was set, more aggressive treatment may be available.
Playing with aperture will confirm or reject the thought that you are getting some perceptible diffraction effects on point lights.
In the days of film, some serious dodging and burning might be invited.
How good is your tripod?
Have you trialed the tripod / camera / timer setup on known perfect point sources with perfect focusing and time delays of the same magnitude? In recent tests at extreme settings and using a focus magnifier I was somewhat surprised at how un-solid and apparently solid tripod mount was and how long it took to settle down after last being touched or knocked . That was extreme (1275mm equivalent focal length - the orbital motion of the Moon at the fl means it takes about 20 seconds for its edge to transit the frame!) but still applicable to your situation and the focus magnifier gives you insight into vibration that the standard viewfinder will never give you.
I haven't mentioned lens quality as it's something of an unknowable and there is enough above to explain what you are seeing (or are not seeing) that really needs addressing before a lens improvement will help vastly. I think! Lens experts by all means disagree.
I haven't mentioned atmospherics, which are also a variable feast and not easily knowable. Evening when cool and calm (or cooler and calmer) is generally a good time in that respect.
Image duplicated to minimise scrolling. Remove if desired.
