How to make RASTER from irregular point data without interpolation

I assume you want your irregular point data on a regular raster. In that case, rasterize should work, and the examples in ?rasterize show how. Here is something based on your data

s100 <- matrix(c(267573.9, 2633781, 213.29545, 262224.4, 2633781, 69.78261, 263742.7, 2633781, 51.21951, 259328.4, 2633781, 301.98413, 264109.8, 2633781, 141.72414, 255094.8, 2633781, 88.90244),  ncol=3,  byrow=TRUE)
colnames(s100) <- c('X', 'Y', 'Z')

library(raster)
# set up an 'empty' raster, here via an extent object derived from your data
e <- extent(s100[,1:2])
e <- e + 1000 # add this as all y's are the same

r <- raster(e, ncol=10, nrow=2)
# or r <- raster(xmn=, xmx=,  ...

# you need to provide a function 'fun' for when there are multiple points per cell
x <- rasterize(s100[, 1:2], r, s100[,3], fun=mean)
plot(x)

This worked for me - the solution was to use SpatialPixelsDataFrame with the suggested tolerance argument (0.916421 in your case):

points <- SpatialPoints(s100_ras[,c('x','y')], s100_ras[,c('z')])
pixels <- SpatialPixelsDataFrame(points, tolerance = 0.916421, points@data)
raster <- raster(pixels[,'z'])

though, due to the high tolerance value, the raster doesn't fit very well to the original points. Could be fit much better.