Haha, but my suggestion was serious.  For the first format he proposed,
this is what I think he wants:

#!/usr/bin/perl
use strict;
use POSIX qw(strftime);

while (<>) {
   chomp;
   s/^\s+//;
   s/\s+$//;
   my($mmyy,$day,@values) = split /\s+/;
   my($month,$year) = $mmyy =~ /(\d+?)(\d\d)$/;
   my $hour = 0;
   foreach my $value (@values) {
       my $time_string = strftime "%Y-%m-%d %H:%M:%S", 0, 0, $hour,
               $day, $month-1, $year;
       print qq{<event data="$time_string" value="$value" />\n};
       $hour += 6;
   }
}

It gives this output for his sample line:

<event data="1995-12-01 00:00:00" value="38.23" />
<event data="1995-12-01 06:00:00" value="39.26" />
<event data="1995-12-01 12:00:00" value="41.10" />
<event data="1995-12-01 18:00:00" value="43.82" />
<event data="1995-12-02 00:00:00" value="47.50" />
<event data="1995-12-02 06:00:00" value="45.29" />

It deviates slightly from exactly what he asked for, in that it prints
month and day with a leading 0 if they are less than 10, but I suspect
that is OK.  (And if not, it can be fixed easily).  I am also assuming
all the data is from before 2000, which seems likely.  If not, that can
be fixed easily, but he didn't say how dates past 1999 are recorded in
the original data.

He's going from data in a text form (flat file), to data in a text form
(XML), so I think it arguably makes sense to treat it as much as you can
as a text transformation, and not get into the issues he's getting into
of integers versus floating point.  Transform the text as
straightforwardly as possible into XML, and then, if he wants to also
validate format, he can supply an XML schema that specifies the formats
of the attributes and validate against that.